Datasets:

james-burton

/

aug-text-exps-v3

like 0

KNeighborsClassifier

C2

German Credit Evaluation

For the case under consideration, the probability of C1 being the correct label is only 12.50%, implying that there is an 87.50% chance that C2 is the true label. The decision above was arrived at mainly based on the values of the following variables F1, F8, and F2. Among these top variables, only F1 has a very strong positive impact on the model, increasing the likelihood of C2 prediction. The most important variables decreasing the prediction are F8 and F2 and the remaining two shifting the verdict away from C2 are F9 and F7. F6 and F5 are the lowest-ranked variables, less important to the prediction made here since they have a moderately low positive impact on the model.

{'C2': '87.50%', 'C1': '12.50%'}

{'F1': 'Checking account', 'F8': 'Saving accounts', 'F2': 'Purpose', 'F9': 'Sex', 'F7': 'Duration', 'F3': 'Housing', 'F4': 'Age', 'F6': 'Job', 'F5': 'Credit amount'}

{'F6': 'F1', 'F5': 'F8', 'F9': 'F2', 'F2': 'F9', 'F8': 'F7', 'F4': 'F3', 'F1': 'F4', 'F3': 'F6', 'F7': 'F5'}

{'C1': 'C2', 'C2': 'C1'}

Good Credit

{'C2': 'Good Credit', 'C1': 'Bad Credit'}

DecisionTreeClassifier

C2

Hotel Satisfaction

With a high degree of confidence, close to 100 percent, the classifier's final label choice for the given case is C2 due to the predicted probability distribution between the class labels. Analysis of the attributions of the input features indicates that the most relevant features driving the classification above are F10, F1, F14, and F5, whereas F8 and F2 are shown to have little contribution to the decision. Furthermore, only four of the features have a negative influence, swinging the classifier decision in this case towards the C1 label and they are F1, F4, F8, and F2. However, except for F1, the contribution of the other negative features is very low when compared with the top positive features such as F14, F9, and F5.

{'C1': '0.00%', 'C2': '100.00%'}

{'F10': 'Type of Travel', 'F1': 'Hotel wifi service', 'F14': 'Other service', 'F5': 'Type Of Booking', 'F9': 'Checkin\\/Checkout service', 'F13': 'Age', 'F12': 'purpose_of_travel', 'F11': 'Common Room entertainment', 'F6': 'Food and drink', 'F4': 'Stay comfort', 'F15': 'Hotel location', 'F7': 'Departure\\/Arrival convenience', 'F3': 'Gender', 'F8': 'Ease of Online booking', 'F2': 'Cleanliness'}

{'F3': 'F10', 'F6': 'F1', 'F14': 'F14', 'F4': 'F5', 'F13': 'F9', 'F5': 'F13', 'F2': 'F12', 'F12': 'F11', 'F10': 'F6', 'F11': 'F4', 'F9': 'F15', 'F7': 'F7', 'F1': 'F3', 'F8': 'F8', 'F15': 'F2'}

{'C2': 'C1', 'C1': 'C2'}

satisfied

{'C1': 'dissatisfied', 'C2': 'satisfied'}

KNNClassifier

C2

Car Acceptability Valuation

Based on the values of the six input features, the model assigned the label C2 to the given case with a higher degree of confidence and according to the model used here, there is a near-zero chance that the label could be C1. Influencing the prediction assessment above are the top four features, F3, F2, and F1, whereas, the least significant feature here is F5. Among the input features, only two, F2 and F5, contradict the label assignment decision above since their values are shifting the label decision in the C1 direction. However, the joint attribution of these features is outweighed by the remaining four features, F3, F1, F4, and F6. This could explain why the model is very certain about the C2 prediction made for the case under consideration.

{'C1': '0.00%', 'C2': '100.00%'}

{'F3': 'persons', 'F2': 'buying', 'F1': 'lug_boot', 'F4': 'maint', 'F6': 'safety', 'F5': 'doors'}

{'F4': 'F3', 'F1': 'F2', 'F5': 'F1', 'F2': 'F4', 'F6': 'F6', 'F3': 'F5'}

{'C1': 'C1', 'C2': 'C2'}

Acceptable

{'C1': 'Unacceptable', 'C2': 'Acceptable'}

RandomForestClassifier

C2

Printer Sales

The most likely label for the given data is C2 and this decision is as the result of the variables passed to the classifier. F7, F18, F22, and F13 are the primary contributors to the aforementioned prediction output. F21, F9, F4, F20, F24, and F3, on the other hand, make insignificant contributions to the classifier labelling the given example. F14 and F19, as well as F16, F25, have a moderate influence on the label selection. The classifier's confidence in the label decision above might be explained away by comparing the greater positive attributions of F19, F14, F7, and F13 to the negative attributions of F16, F6, F22, F11, F18, and F15.

{'C1': '20.00%', 'C2': '80.00%'}

{'F13': 'X24', 'F7': 'X1', 'F22': 'X8', 'F18': 'X21', 'F16': 'X4', 'F19': 'X10', 'F14': 'X3', 'F25': 'X15', 'F17': 'X9', 'F10': 'X23', 'F23': 'X25', 'F1': 'X7', 'F11': 'X22', 'F2': 'X11', 'F6': 'X17', 'F5': 'X18', 'F8': 'X26', 'F15': 'X13', 'F26': 'X6', 'F12': 'X20', 'F20': 'X16', 'F3': 'X19', 'F4': 'X2', 'F21': 'X12', 'F24': 'X5', 'F9': 'X14'}

{'F24': 'F13', 'F1': 'F7', 'F8': 'F22', 'F21': 'F18', 'F4': 'F16', 'F10': 'F19', 'F3': 'F14', 'F15': 'F25', 'F9': 'F17', 'F23': 'F10', 'F25': 'F23', 'F7': 'F1', 'F22': 'F11', 'F11': 'F2', 'F17': 'F6', 'F18': 'F5', 'F26': 'F8', 'F13': 'F15', 'F6': 'F26', 'F20': 'F12', 'F16': 'F20', 'F19': 'F3', 'F2': 'F4', 'F12': 'F21', 'F5': 'F24', 'F14': 'F9'}

{'C2': 'C1', 'C1': 'C2'}

More

{'C1': 'Less', 'C2': 'More'}

LogisticRegression

C4

Flight Price-Range Classification

Because the prediction algorithm outputs reveal that the likelihood of C4 being the correct label is equal to 93.02%; hence, there is only a little possibility that the true label for the provided data instance is either of the other labels, C1, C2, and C3. The variables F5, F7, F3, and F4 are the most crucial ones driving the label assignment conclusion above, whereas F10, F11, and F9 are the least vital ones. Taking into account the direction of effect of each input feature, as demonstrated by the attribution analysis, it is possible to deduce that the positive features driving the prediction upward towards C4 are F5, F6, F3, F1, F4, F7, and F11. The negative contributions of F8, F2, F9, F10, and F12 are ascribed to the marginal uncertainty in the expected output decision. When the predicted probabilities across the classes are considered, it is possible to infer that the combined positive contribution outranks the negative contributions; therefore, the algorithm is certain that C4 is the real label.

{'C4': '93.02%', 'C2': '6.97%', 'C3': '0.01%', 'C1': '0.0%'}

{'F5': 'Total_Stops', 'F7': 'Airline', 'F4': 'Destination', 'F3': 'Journey_day', 'F8': 'Source', 'F2': 'Dep_hour', 'F1': 'Duration_hours', 'F12': 'Dep_minute', 'F6': 'Duration_mins', 'F10': 'Arrival_minute', 'F11': 'Arrival_hour', 'F9': 'Journey_month'}

{'F12': 'F5', 'F9': 'F7', 'F11': 'F4', 'F1': 'F3', 'F10': 'F8', 'F3': 'F2', 'F7': 'F1', 'F4': 'F12', 'F8': 'F6', 'F6': 'F10', 'F5': 'F11', 'F2': 'F9'}

{'C1': 'C4', 'C2': 'C2', 'C4': 'C3', 'C3': 'C1'}

Low

{'C4': 'Low', 'C2': 'Moderate', 'C3': 'High', 'C1': 'Special'}

LogisticRegression

C2

Food Ordering Customer Churn Prediction

Based mainly on the values of the input variables F4, F32, F9, and F38, the predictor classifies the case as C2 with a 90.15% labelling confidence level, indicating that there is only a 9.85% probability that the right label could be C1. Variables that contribute positively to the prediction verdict include F4, F26, F19, and F38. The values of these variables increase the odds of the model labelling the given case as C2. On the other hand, F9, F32, F39, and F30 are the variables influencing the prediction decision in favour of C1 instead of C2. Simply put, the values of these negative variables contradict the label assigned here and finally, the model places little emphasis on the values of features such as F27, F34, F18, and F35 when determining the correct label in this instance, as they have nearly zero influence.

{'C1': '9.85%', 'C2': '90.15%'}

{'F4': 'Unaffordable', 'F9': 'Perference(P2)', 'F32': 'Influence of rating', 'F38': 'Good Food quality', 'F19': 'Delay of delivery person picking up food', 'F26': 'Less Delivery time', 'F39': 'Freshness ', 'F16': 'Politeness', 'F30': 'Ease and convenient', 'F31': 'More restaurant choices', 'F44': 'Missing item', 'F24': 'Order Time', 'F22': 'Gender', 'F6': 'Time saving', 'F13': 'Unavailability', 'F3': 'Late Delivery', 'F23': 'Temperature', 'F46': 'High Quality of package', 'F33': 'Long delivery time', 'F11': 'Poor Hygiene', 'F34': 'Low quantity low time', 'F27': 'Delivery person ability', 'F18': 'Number of calls', 'F35': 'Google Maps Accuracy', 'F10': 'Residence in busy location', 'F41': 'Good Taste ', 'F37': 'Maximum wait time', 'F15': 'Influence of time', 'F45': 'Good Road Condition', 'F12': 'Age', 'F43': 'Order placed by mistake', 'F42': 'Wrong order delivered', 'F7': 'Delay of delivery person getting assigned', 'F21': 'Family size', 'F5': 'Bad past experience', 'F1': 'Health Concern', 'F25': 'Self Cooking', 'F14': 'Good Tracking system', 'F29': 'More Offers and Discount', 'F8': 'Easy Payment option', 'F20': 'Perference(P1)', 'F17': 'Educational Qualifications', 'F36': 'Monthly Income', 'F40': 'Occupation', 'F2': 'Marital Status', 'F28': 'Good Quantity'}

{'F23': 'F4', 'F9': 'F9', 'F38': 'F32', 'F15': 'F38', 'F26': 'F19', 'F39': 'F26', 'F43': 'F39', 'F42': 'F16', 'F10': 'F30', 'F12': 'F31', 'F28': 'F44', 'F31': 'F24', 'F2': 'F22', 'F11': 'F6', 'F22': 'F13', 'F19': 'F3', 'F44': 'F23', 'F40': 'F46', 'F24': 'F33', 'F20': 'F11', 'F36': 'F34', 'F37': 'F27', 'F41': 'F18', 'F34': 'F35', 'F33': 'F10', 'F45': 'F41', 'F32': 'F37', 'F30': 'F15', 'F35': 'F45', 'F1': 'F12', 'F29': 'F43', 'F27': 'F42', 'F25': 'F7', 'F7': 'F21', 'F21': 'F5', 'F18': 'F1', 'F17': 'F25', 'F16': 'F14', 'F14': 'F29', 'F13': 'F8', 'F8': 'F20', 'F6': 'F17', 'F5': 'F36', 'F4': 'F40', 'F3': 'F2', 'F46': 'F28'}

{'C2': 'C1', 'C1': 'C2'}

Go Away

{'C1': 'Return', 'C2': 'Go Away'}

LogisticRegression

C3

Air Quality Prediction

The classification output decision is based solely on the information supplied to the model and it predicts class C3 with a higher confidence level, equal to 94.10%, indicating the model is very confident that the correct label for the given case is not either class C4 or class C1 or class C2. The classification output decision with regards to the given case boils down to the values of the features F3, F6, F4, and F1, which are shown to have the most significant influence on the model. Among these relevant features, only F3, F1, and F6 have a positive impact, increasing the response towards labelling the case as C3. Conversely, the remaining ones, F4 and F5, have negative attributions, decreasing the odds of the assigned label. Finally, feature F2 has little impact on this prediction among the features since its value received little consideration from the model.

{'C4': '0.00%', 'C1': '0.53%', 'C3': '94.10%', 'C2': '5.37%'}

{'F3': 'MQ5', 'F6': 'MQ6', 'F1': 'MQ3', 'F4': 'MQ4', 'F5': 'MQ1', 'F2': 'MQ2'}

{'F5': 'F3', 'F6': 'F6', 'F3': 'F1', 'F4': 'F4', 'F1': 'F5', 'F2': 'F2'}

{'C3': 'C4', 'C1': 'C1', 'C4': 'C3', 'C2': 'C2'}

Cleaning

{'C4': 'Preparing meals', 'C1': 'Presence of smoke', 'C3': 'Cleaning', 'C2': 'Other'}

SVC

C1

German Credit Evaluation

The model predicts that this case is likely C1 with a confidence level equal to 66.80%, meaning there is a 33.20% chance that it could be C2 instead. According to the analysis for this case under consideration, the most relevant features considered by the model are F2, F6, F9, F4, and F8, however, the least relevant features are F3 and F5. The F9, F4, F8, and F7 can be regarded as positively supporting features given that they increase the model's response in favour of the prediction conclusion above. In contrast, the F2, F6, and F1 are the features supporting the prediction of the alternative or other class label C2. Even though only a small number of features support the prediction of C2, their collective or joint influence is enough to upset the joint influence of the other features, leading to the uncertainty of the C1 prediction.

{'C1': '66.80%', 'C2': '33.20%'}

{'F2': 'Saving accounts', 'F6': 'Duration', 'F9': 'Checking account', 'F4': 'Sex', 'F8': 'Age', 'F7': 'Purpose', 'F1': 'Housing', 'F3': 'Job', 'F5': 'Credit amount'}

{'F5': 'F2', 'F8': 'F6', 'F6': 'F9', 'F2': 'F4', 'F1': 'F8', 'F9': 'F7', 'F4': 'F1', 'F3': 'F3', 'F7': 'F5'}

{'C2': 'C1', 'C1': 'C2'}

Good Credit

{'C1': 'Good Credit', 'C2': 'Bad Credit'}

SGDClassifier

C2

House Price Classification

Based on the values of the input variables resulting in the predicted likelihoods across the classes, the classification algorithm is confident that the right label for the provided data is C2. According to the algorithm, there is no possibility that C1 is the correct label. However, the attributions of F10, F12, F4, and F1 indicate that the correct label might be C1 rather than C2. The top four variables are F11, F5, F6, and F7, all of which have a positive influence on the algorithm's prediction output, hence confirming the C2 classification. This conclusion is further supported by the contributions of F3, F9, F8, F13, and F2, which are also positive variables.

{'C1': '0.0%', 'C2': '100.0%'}

{'F11': 'AGE', 'F5': 'RAD', 'F6': 'LSTAT', 'F7': 'RM', 'F10': 'DIS', 'F12': 'CHAS', 'F3': 'ZN', 'F9': 'CRIM', 'F4': 'TAX', 'F2': 'B', 'F13': 'PTRATIO', 'F8': 'INDUS', 'F1': 'NOX'}

{'F7': 'F11', 'F9': 'F5', 'F13': 'F6', 'F6': 'F7', 'F8': 'F10', 'F4': 'F12', 'F2': 'F3', 'F1': 'F9', 'F10': 'F4', 'F12': 'F2', 'F11': 'F13', 'F3': 'F8', 'F5': 'F1'}

{'C2': 'C1', 'C1': 'C2'}

High

{'C1': 'Low', 'C2': 'High'}

KNeighborsClassifier

C2

Suspicious Bidding Identification

With a higher degree of confidence, the classifier assigns the label C2 due to the fact that there is a close to zero chance that C1 is the label. The confidence level with respect to this classification output is largely due to the strong positive influence of F4. However, decreasing the probability that C2 is the true label are the negative features F9, F8, F7, F2, F6, and F1. Furthermore, F3 and F5 also increase the likelihood of C2 being the true label. In conclusion, the joint impact of the negative features is very weak compared to the positive features, hence the strong driving force of the classifier to assign the chosen label, C2.

{'C2': '99.90%', 'C1': '0.10%'}

{'F4': 'Z3', 'F9': 'Z9', 'F8': 'Z8', 'F7': 'Z1', 'F5': 'Z5', 'F3': 'Z4', 'F2': 'Z2', 'F6': 'Z6', 'F1': 'Z7'}

{'F3': 'F4', 'F9': 'F9', 'F8': 'F8', 'F1': 'F7', 'F5': 'F5', 'F4': 'F3', 'F2': 'F2', 'F6': 'F6', 'F7': 'F1'}

{'C2': 'C2', 'C1': 'C1'}

Normal

{'C2': 'Normal', 'C1': 'Suspicious'}

MLPClassifier

C1

Annual Income Earnings

According to the input variables, there is a 99.81% chance that C1 is the correct label for the given data instance, with a prediction probability of the alternative label, C2, equal to 0.19% which shows that there is little chance that C2 is the true label. F5, F1, and F8 are the top contributing features leading to the classification decision here. On the contrary, the F10, F13, and F4 are the least relevant features. The input features with moderate influence are F11, F2, F7, F12, F3, F14, and F6. Even though the different features have some level of influence on the classification, not all of them positively contribute. Actually, F7, F9, F6, and F13 have negative attributions, decreasing the classifier's response towards assigning C1; however, the joint influence of these features is outweighed by the positive attributions of F5, F1, F8, F11, and F2.

{'C1': '99.81%', 'C2': '0.19%'}

{'F5': 'Capital Gain', 'F1': 'Marital Status', 'F8': 'Capital Loss', 'F11': 'Age', 'F2': 'Hours per week', 'F12': 'Education', 'F7': 'Occupation', 'F9': 'Country', 'F3': 'Relationship', 'F14': 'Workclass', 'F6': 'Sex', 'F10': 'fnlwgt', 'F13': 'Education-Num', 'F4': 'Race'}

{'F11': 'F5', 'F6': 'F1', 'F12': 'F8', 'F1': 'F11', 'F13': 'F2', 'F4': 'F12', 'F7': 'F7', 'F14': 'F9', 'F8': 'F3', 'F2': 'F14', 'F10': 'F6', 'F3': 'F10', 'F5': 'F13', 'F9': 'F4'}

{'C2': 'C1', 'C1': 'C2'}

Under 50K

{'C1': 'Under 50K', 'C2': 'Above 50K'}

SVC

C2

Australian Credit Approval

With a confidence level equal to 81.43%, the classification algorithm labels the given data as C2, however, there is about an 18.57% chance that C1 could be the right label. The assignment of C2 to the given case is mainly based on the positive influence and contribution of input features F7, F6, and F9. Furthermore, the majority of the remaining input features have positive contributions, further increasing the predictability of label C2. F8, F14, F4, and F3 are the features with negative contributions, shifting the decision towards C1 instead of C2. Summarizing, comparing the attributions of the negative features to even those of the top three positive features explains why the algorithm is certain that C2 is the right label here.

{'C1': '18.57%', 'C2': '81.43%'}

{'F7': 'A8', 'F6': 'A9', 'F9': 'A14', 'F13': 'A12', 'F11': 'A7', 'F1': 'A4', 'F10': 'A5', 'F2': 'A11', 'F8': 'A1', 'F14': 'A13', 'F5': 'A10', 'F4': 'A2', 'F12': 'A6', 'F3': 'A3'}

{'F8': 'F7', 'F9': 'F6', 'F14': 'F9', 'F12': 'F13', 'F7': 'F11', 'F4': 'F1', 'F5': 'F10', 'F11': 'F2', 'F1': 'F8', 'F13': 'F14', 'F10': 'F5', 'F2': 'F4', 'F6': 'F12', 'F3': 'F3'}

{'C1': 'C1', 'C2': 'C2'}

Class 2

{'C1': 'Class 1', 'C2': 'Class 2'}

DNN

C2

Broadband Sevice Signup

The classification model employed here is very certain that the correct label is C2, implying that there is a near-zero chance that C1 is the label. The top six variables with the most influence on the prediction are all shifting the prediction in favour of C2. This might explain why the model is very certain about the C2 label and these top positive attributes are F8, F32, F26, F27, F15, and F19. F29, F38, F9, F42, and F40 all have moderately low-negative contributions, weakly swinging the direction of the model's decision towards C1. Finally, the decision to label the case as C2 is marginally supported by F33 and F11, whereas F30, F37, and F20 suggest that C1 could be the true label. In conclusion, the very high confidence level with regard to this prediction can be explained away by considering the very strong positive influence of F8, F32, F27, and F26.

{'C1': '0.00%', 'C2': '100.00%'}

{'F8': 'X38', 'F32': 'X5', 'F26': 'X33', 'F27': 'X37', 'F15': 'X27', 'F19': 'X3', 'F39': 'X16', 'F29': 'X41', 'F38': 'X2', 'F9': 'X39', 'F42': 'X29', 'F23': 'X25', 'F34': 'X1', 'F13': 'X19', 'F18': 'X10', 'F24': 'X18', 'F40': 'X26', 'F3': 'X35', 'F1': 'X40', 'F6': 'X24', 'F35': 'X32', 'F22': 'X22', 'F21': 'X21', 'F4': 'X6', 'F10': 'X14', 'F2': 'X42', 'F41': 'X30', 'F17': 'X28', 'F5': 'X34', 'F25': 'X23', 'F16': 'X9', 'F31': 'X20', 'F12': 'X11', 'F7': 'X12', 'F14': 'X8', 'F33': 'X15', 'F11': 'X31', 'F30': 'X17', 'F37': 'X13', 'F20': 'X7', 'F28': 'X36', 'F36': 'X4'}

{'F35': 'F8', 'F41': 'F32', 'F30': 'F26', 'F34': 'F27', 'F25': 'F15', 'F2': 'F19', 'F14': 'F39', 'F39': 'F29', 'F1': 'F38', 'F36': 'F9', 'F42': 'F42', 'F23': 'F23', 'F40': 'F34', 'F17': 'F13', 'F8': 'F18', 'F16': 'F24', 'F24': 'F40', 'F32': 'F3', 'F37': 'F1', 'F22': 'F6', 'F29': 'F35', 'F20': 'F22', 'F19': 'F21', 'F4': 'F4', 'F12': 'F10', 'F38': 'F2', 'F27': 'F41', 'F26': 'F17', 'F31': 'F5', 'F21': 'F25', 'F7': 'F16', 'F18': 'F31', 'F9': 'F12', 'F10': 'F7', 'F6': 'F14', 'F13': 'F33', 'F28': 'F11', 'F15': 'F30', 'F11': 'F37', 'F5': 'F20', 'F33': 'F28', 'F3': 'F36'}

{'C2': 'C1', 'C1': 'C2'}

Yes

{'C1': 'No', 'C2': 'Yes'}

KNeighborsClassifier

C2

Water Quality Classification

For the given case, the model generates the label C2 instead of C1, since C2 has a higher prediction likelihood than C1. According to the attribution graph shown, F8, and F6 are the most influential variables, resulting in the classification verdict above. F2, F9, and F3, on the other hand, are the least important variables considered by the model. F4, F1, F5, and F7 are shown to have a moderate influence on the classification made here. To sum up, with F2, F9, and F3 being the only variables contributing negatively, it is foreseeable why the model is quite certain that C1 is not the correct label for the given case.

{'C2': '87.50%', 'C1': '12.50%'}

{'F8': 'Hardness', 'F6': 'Sulfate', 'F4': 'Solids', 'F1': 'ph', 'F5': 'Organic_carbon', 'F7': 'Conductivity', 'F2': 'Trihalomethanes', 'F9': 'Turbidity', 'F3': 'Chloramines'}

{'F2': 'F8', 'F5': 'F6', 'F3': 'F4', 'F1': 'F1', 'F7': 'F5', 'F6': 'F7', 'F8': 'F2', 'F9': 'F9', 'F4': 'F3'}

{'C2': 'C2', 'C1': 'C1'}

Not Portable

{'C2': 'Not Portable', 'C1': 'Portable'}

DecisionTreeClassifier

C2

Concrete Strength Classification

Per the classifier, the most probable class with a very high confidence level is C2 mainly because the probability that C1 is the correct label is zero. From the attributions analysis, all the inputs are shown to contribute to or influence the above classification. The ranking of the features from the least important to the most important based on their degree of influence is as follows: F7, F8, F6, F2, F3, F4, F1, F5. Simply looking at the attributions of the input features, it is obvious why the classifier is very confident that C1 is not the correct label for the given All the features have positive contributions, resulting in a strong push towards C2.

{'C2': '100.00%', 'C1': '0.00%'}

{'F5': 'age_days', 'F1': 'superplasticizer', 'F4': 'cement', 'F3': 'coarseaggregate', 'F2': 'fineaggregate', 'F6': 'water', 'F8': 'slag', 'F7': 'flyash'}

{'F8': 'F5', 'F5': 'F1', 'F1': 'F4', 'F6': 'F3', 'F7': 'F2', 'F4': 'F6', 'F2': 'F8', 'F3': 'F7'}

{'C1': 'C2', 'C2': 'C1'}

Weak

{'C2': 'Weak', 'C1': 'Strong'}

KNeighborsClassifier

C1

Air Quality Prediction

The model predicted the C1 class for the test case with a very high degree of confidence. F6 is the only feature contributing against the prediction of the C1 class, while F3 and F5 contributed positively towards the prediction of C1. In decreasing order, F4, F2 and F1 were the three features with the least positive impact on the prediction of C1. Overall, given that only F6 has negative influence on the decision, it is not surprising to see the associated confidence level of the assigned label.

{'C4': '0.00%', 'C1': '100.00%', 'C2': '0.00%', 'C3': '0.00%'}

{'F6': 'MQ6', 'F3': 'MQ4', 'F5': 'MQ5', 'F4': 'MQ2', 'F2': 'MQ1', 'F1': 'MQ3'}

{'F6': 'F6', 'F4': 'F3', 'F5': 'F5', 'F2': 'F4', 'F1': 'F2', 'F3': 'F1'}

{'C2': 'C4', 'C3': 'C1', 'C1': 'C2', 'C4': 'C3'}

Presence of smoke

{'C4': 'Preparing meals', 'C1': 'Presence of smoke', 'C2': 'Cleaning', 'C3': 'Other'}

SVM_linear

C1

Mobile Price-Range Classification

Per the classification algorithm, the most probable class is C1 since the prediction probabilities indicate there is little to no chance that the correct label for the given data instance is any of the following classes: C4, C3, and C2. This labelling is primarily owing to the roles that the features F4, F1, and F7 performed. On the lower end of the spectrum are the input features F15, F2, F19, and F20, which are demonstrated to be less essential for this labelling assignment task. Finally, only F8 and F16 are features having a negative effect, reducing the likelihood of C1 being the accurate classification here.

{'C4': '0.00%', 'C3': '0.00%', 'C2': '0.00%', 'C1': '100.00%'}

{'F1': 'ram', 'F7': 'battery_power', 'F4': 'px_width', 'F16': 'int_memory', 'F8': 'sc_h', 'F10': 'pc', 'F18': 'mobile_wt', 'F3': 'fc', 'F6': 'n_cores', 'F17': 'clock_speed', 'F13': 'blue', 'F14': 'three_g', 'F12': 'touch_screen', 'F5': 'm_dep', 'F11': 'px_height', 'F9': 'talk_time', 'F19': 'dual_sim', 'F20': 'wifi', 'F15': 'four_g', 'F2': 'sc_w'}

{'F11': 'F1', 'F1': 'F7', 'F10': 'F4', 'F4': 'F16', 'F12': 'F8', 'F8': 'F10', 'F6': 'F18', 'F3': 'F3', 'F7': 'F6', 'F2': 'F17', 'F15': 'F13', 'F18': 'F14', 'F19': 'F12', 'F5': 'F5', 'F9': 'F11', 'F14': 'F9', 'F16': 'F19', 'F20': 'F20', 'F17': 'F15', 'F13': 'F2'}

{'C1': 'C4', 'C3': 'C3', 'C4': 'C2', 'C2': 'C1'}

r4

{'C4': 'r1', 'C3': 'r2', 'C2': 'r3', 'C1': 'r4'}

GradientBoostingClassifier

C1

Australian Credit Approval

The predicted label is C1 at a confidence level of 92.11%, insinuating that there is a 7.89% chance that the label could be C2. In this case, the feature with the most significant influence on the model's decision is F4, with a very strong positive contribution in support of the C1 prediction. The next set of features with moderately high impact is F3, F1, F2, F5, and F14. Among this set, only F14 and F3 have a negative influence in support of label C2. Finally, on the lower end, the values of F13, F11, and F6 are deemed less important by the model when labelling this case.

{'C2': '7.89%', 'C1': '92.11%'}

{'F4': 'A8', 'F3': 'A14', 'F1': 'A4', 'F2': 'A7', 'F5': 'A13', 'F14': 'A9', 'F12': 'A2', 'F10': 'A3', 'F8': 'A10', 'F9': 'A5', 'F7': 'A1', 'F13': 'A11', 'F11': 'A12', 'F6': 'A6'}

{'F8': 'F4', 'F14': 'F3', 'F4': 'F1', 'F7': 'F2', 'F13': 'F5', 'F9': 'F14', 'F2': 'F12', 'F3': 'F10', 'F10': 'F8', 'F5': 'F9', 'F1': 'F7', 'F11': 'F13', 'F12': 'F11', 'F6': 'F6'}

{'C2': 'C2', 'C1': 'C1'}

Class 2

{'C2': 'Class 1', 'C1': 'Class 2'}

SVM

C1

Customer Churn Modelling

Taking into account the values of the input features, the prediction model's output for the case under consideration is C1. Given that there is a 27.27% probability that it could be C2, this labelling decision is not 100.0% certain. For the case under consideration, the label assignment is mainly due to the values of F5, F2, F4, and F9. F9 is identified as the most important or relevant, while F7 is considered the least important, since its contribution to the model is only marginal. In terms of the influence direction of each feature, F9 and F2 have a very strong positive contribution, driving the prediction higher toward the C1 class followed by F5, F4, and F1 all with moderately positive influence, whereas F7 has a negligible positive impact on the model in this case. Finally, for this case, F8, F3, F10, and F6 all have a negative impact on the prediction verdict, however, their pull or influence is not enough to transfer predictions in the direction of another class label, C2.

{'C2': '27.27%', 'C1': '72.73%'}

{'F9': 'Age', 'F2': 'IsActiveMember', 'F5': 'Geography', 'F4': 'NumOfProducts', 'F1': 'Gender', 'F8': 'Tenure', 'F3': 'CreditScore', 'F10': 'EstimatedSalary', 'F6': 'Balance', 'F7': 'HasCrCard'}

{'F4': 'F9', 'F9': 'F2', 'F2': 'F5', 'F7': 'F4', 'F3': 'F1', 'F5': 'F8', 'F1': 'F3', 'F10': 'F10', 'F6': 'F6', 'F8': 'F7'}

{'C1': 'C2', 'C2': 'C1'}

Leave

{'C2': 'Stay', 'C1': 'Leave'}

SVMClassifier_liner

C2

Employee Attrition

The prediction output decision by the model is that the likelihood of label C2 is 94.15% and that of class C1 is only around 5.85%, meaning the model is certain that C2 is likely the true label for the given case. First of all, the classification is performed with negligible contributions from the variables F18, F4, F5, F2, and F13 since their attributions are very close to zero. However, examination or inspection of the attributions of the different variables reveals that F6, F22, F29, F1, and F24 are the highly influential ones driving the predicted probabilities across the classes. In addition, the decision about the correct label for this case is moderately influenced by the values of F16, F9, F25, F19, F10, F20, and F14. In terms of the direction of influence or contributions of the variables, F6, F29, F1, F9, and F25 are the top positive variables, encouraging the predicted output to be equal to C2. Pushing the decision towards the C2 label and further away from C1 are the contriutions of the variables F10, F20, F27, and F30. Finally, the 5.85% likelihood of C1 can be attributed to the negative contributions of the top negative variables F22, F24, F30, F19, and F14.

{'C2': '94.15%', 'C1': '5.85%'}

{'F6': 'OverTime', 'F22': 'MaritalStatus', 'F29': 'NumCompaniesWorked', 'F1': 'BusinessTravel', 'F24': 'TotalWorkingYears', 'F16': 'DistanceFromHome', 'F9': 'YearsSinceLastPromotion', 'F25': 'Department', 'F19': 'Gender', 'F10': 'EnvironmentSatisfaction', 'F20': 'PerformanceRating', 'F14': 'Education', 'F27': 'JobRole', 'F30': 'YearsAtCompany', 'F8': 'JobInvolvement', 'F23': 'EducationField', 'F12': 'JobSatisfaction', 'F17': 'TrainingTimesLastYear', 'F26': 'HourlyRate', 'F28': 'WorkLifeBalance', 'F7': 'Age', 'F18': 'RelationshipSatisfaction', 'F4': 'DailyRate', 'F5': 'YearsInCurrentRole', 'F13': 'StockOptionLevel', 'F2': 'PercentSalaryHike', 'F3': 'MonthlyRate', 'F11': 'MonthlyIncome', 'F15': 'JobLevel', 'F21': 'YearsWithCurrManager'}

{'F26': 'F6', 'F25': 'F22', 'F8': 'F29', 'F17': 'F1', 'F11': 'F24', 'F3': 'F16', 'F15': 'F9', 'F21': 'F25', 'F23': 'F19', 'F28': 'F10', 'F19': 'F20', 'F27': 'F14', 'F24': 'F27', 'F13': 'F30', 'F29': 'F8', 'F22': 'F23', 'F30': 'F12', 'F12': 'F17', 'F4': 'F26', 'F20': 'F28', 'F1': 'F7', 'F18': 'F18', 'F2': 'F4', 'F14': 'F5', 'F10': 'F13', 'F9': 'F2', 'F7': 'F3', 'F6': 'F11', 'F5': 'F15', 'F16': 'F21'}

{'C2': 'C2', 'C1': 'C1'}

Stay

{'C2': 'Leave', 'C1': 'Leave'}

RandomForestClassifier

C2

Printer Sales

The most probable label, according to the classifier for the given data, is C2, which happens to have a higher predicted probability than that of C1. The major players in the above prediction output are F4, F13, F25, and F10. Conversely, F19, F3, F9, F22, F26, and F11 have negligible contributions when it comes to the classifier labelling the given case. Features such as F21, F7, F20, and F1 have a moderate influence on the decision. Comparing the stronger positive attributions of F13, F4, F7, and F20 to the negative attributions of F25, F10, F21, F2, F17, and F24 could explain why the classifier is quite confident in the label choice above.

{'C1': '20.00%', 'C2': '80.00%'}

{'F4': 'X24', 'F13': 'X1', 'F25': 'X8', 'F10': 'X21', 'F21': 'X4', 'F7': 'X10', 'F20': 'X3', 'F1': 'X15', 'F8': 'X9', 'F14': 'X23', 'F16': 'X25', 'F12': 'X7', 'F2': 'X22', 'F5': 'X11', 'F17': 'X17', 'F18': 'X18', 'F23': 'X26', 'F24': 'X13', 'F6': 'X6', 'F15': 'X20', 'F19': 'X16', 'F3': 'X19', 'F9': 'X2', 'F22': 'X12', 'F26': 'X5', 'F11': 'X14'}

{'F24': 'F4', 'F1': 'F13', 'F8': 'F25', 'F21': 'F10', 'F4': 'F21', 'F10': 'F7', 'F3': 'F20', 'F15': 'F1', 'F9': 'F8', 'F23': 'F14', 'F25': 'F16', 'F7': 'F12', 'F22': 'F2', 'F11': 'F5', 'F17': 'F17', 'F18': 'F18', 'F26': 'F23', 'F13': 'F24', 'F6': 'F6', 'F20': 'F15', 'F16': 'F19', 'F19': 'F3', 'F2': 'F9', 'F12': 'F22', 'F5': 'F26', 'F14': 'F11'}

{'C2': 'C1', 'C1': 'C2'}

More

{'C1': 'Less', 'C2': 'More'}

AdaBoostClassifier

C1

Air Quality Prediction

The most likely label is C1 since there is a 30.83% chance it could be C2, a 35.74% chance it could be C1, and a 33.42% chance it could be C4. Therefore, the correct label is not C3, which the model is very certain about. The above decision is primarily controlled by the values F1, F2, F4, and F3 which are shown to have positive influences that support the model's classification judgement here. In contrast, the remaining features F5 and F6 negatively support the classification decision, decreasing the chances of C1 being the correct label. In view of the fact that the probability distributions across the classes, we can conclude that the model is very uncertain about which label is appropriate for the given data instance and the features F5 and F6 should be blamed for this.

{'C2': '30.83%', 'C1': '35.74%', 'C3': '0.00%', 'C4': '33.42%'}

{'F1': 'MQ5', 'F2': 'MQ3', 'F4': 'MQ2', 'F3': 'MQ6', 'F5': 'MQ1', 'F6': 'MQ4'}

{'F5': 'F1', 'F3': 'F2', 'F2': 'F4', 'F6': 'F3', 'F1': 'F5', 'F4': 'F6'}

{'C1': 'C2', 'C4': 'C1', 'C2': 'C3', 'C3': 'C4'}

Presence of smoke

{'C2': 'Preparing meals', 'C1': 'Presence of smoke', 'C3': 'Cleaning', 'C4': 'Other'}

KNeighborsClassifier

C1

Ethereum Fraud Detection

Because the prediction probability of C2 is equal to 0.0%, the presented case is labelled as C1 with a very high level of confidence. For this classification scenario, the input features that have the greatest influence on the end outcome are F22, F20, F21, and F12. F23, F11, F2, F31, and F3 have a mild impact. However, because F37, F15, F17, and F8 have insignificant attribution values, they have little influence on the model's judgement. Among the top features, F22, F20, F21, and F12, only F22 and F21 exhibit negative attributions that favour the least likely class, C2, whereas F20 and F12 positively support the model's classification result for the provided data. Finally, only F1 and F7 positively contribute to the model's decision among the remaining significant features: F7, F34, F16, F1, F33, F6, and F5.

{'C2': '0.00%', 'C1': '100.00%'}

{'F22': 'Time Diff between first and last (Mins)', 'F20': 'Unique Received From Addresses', 'F21': 'Avg min between received tnx', 'F12': 'min val sent', 'F23': ' ERC20 min val rec', 'F11': 'Sent tnx', 'F2': 'min value received', 'F31': 'avg val sent', 'F3': ' ERC20 uniq rec addr', 'F29': ' ERC20 avg val sent', 'F10': ' ERC20 uniq rec contract addr', 'F25': ' ERC20 uniq rec token name', 'F9': 'max val sent', 'F7': 'Unique Sent To Addresses', 'F16': 'total transactions (including tnx to create contract', 'F34': 'avg val received', 'F1': ' ERC20 uniq sent addr.1', 'F33': ' ERC20 uniq sent token name', 'F6': ' Total ERC20 tnxs', 'F5': 'Received Tnx', 'F37': ' ERC20 uniq sent addr', 'F15': ' ERC20 max val sent', 'F17': ' ERC20 min val sent', 'F8': ' ERC20 avg val rec', 'F28': ' ERC20 max val rec', 'F14': 'Avg min between sent tnx', 'F35': ' ERC20 total Ether sent contract', 'F26': ' ERC20 total ether sent', 'F27': ' ERC20 total Ether received', 'F36': 'total ether balance', 'F30': 'total ether sent contracts', 'F32': 'total Ether sent', 'F38': 'avg value sent to contract', 'F24': 'max val sent to contract', 'F13': 'min value sent to contract', 'F19': 'max value received ', 'F4': 'Number of Created Contracts', 'F18': 'total ether received'}

{'F3': 'F22', 'F7': 'F20', 'F2': 'F21', 'F12': 'F12', 'F31': 'F23', 'F4': 'F11', 'F9': 'F2', 'F14': 'F31', 'F28': 'F3', 'F36': 'F29', 'F30': 'F10', 'F38': 'F25', 'F13': 'F9', 'F8': 'F7', 'F18': 'F16', 'F11': 'F34', 'F29': 'F1', 'F37': 'F33', 'F23': 'F6', 'F5': 'F5', 'F27': 'F37', 'F35': 'F15', 'F34': 'F17', 'F33': 'F8', 'F32': 'F28', 'F1': 'F14', 'F26': 'F35', 'F25': 'F26', 'F24': 'F27', 'F22': 'F36', 'F21': 'F30', 'F19': 'F32', 'F17': 'F38', 'F16': 'F24', 'F15': 'F13', 'F10': 'F19', 'F6': 'F4', 'F20': 'F18'}

{'C2': 'C2', 'C1': 'C1'}

Fraud

{'C2': 'Not Fraud', 'C1': 'Fraud'}

RandomForestClassifier

C3

Mobile Price-Range Classification

Between the four possible classes, the label for this case is predicted as C3, with a 73.08% likelihood that this is correct. With a likelihood of about 26.92%, the next probable label is shown to be C1. The prediction assessment above is mainly based on the values of the features F6, F18, F8, F3, and F11. The strongest impact came from F6, followed by F8, F18, F11, and F3. The collective contributions of the positive features F6, F18, F4, and F12 far outweigh the contributions of the negative attributes F8, F11, F3, and F9. Of the twenty attributes, majority of them are shown to have values pushing the prediction towards one of the three other possible classes and as such, it is surprising to see that the model is not 100% confident in the C3 prediction. On the grounds that the likelihood of C3 being correct is 73.08%, we can conclude that the model is quite confident with its final decision for the case under consideration.

{'C3': '73.08%', 'C1': '26.92%', 'C4': '0.00%', 'C2': '0.00%'}

{'F6': 'ram', 'F8': 'px_width', 'F18': 'battery_power', 'F11': 'px_height', 'F3': 'n_cores', 'F4': 'dual_sim', 'F12': 'touch_screen', 'F9': 'int_memory', 'F17': 'wifi', 'F10': 'fc', 'F1': 'four_g', 'F19': 'm_dep', 'F15': 'pc', 'F16': 'mobile_wt', 'F14': 'talk_time', 'F5': 'three_g', 'F7': 'sc_h', 'F20': 'sc_w', 'F13': 'blue', 'F2': 'clock_speed'}

{'F11': 'F6', 'F10': 'F8', 'F1': 'F18', 'F9': 'F11', 'F7': 'F3', 'F16': 'F4', 'F19': 'F12', 'F4': 'F9', 'F20': 'F17', 'F3': 'F10', 'F17': 'F1', 'F5': 'F19', 'F8': 'F15', 'F6': 'F16', 'F14': 'F14', 'F18': 'F5', 'F12': 'F7', 'F13': 'F20', 'F15': 'F13', 'F2': 'F2'}

{'C4': 'C3', 'C1': 'C1', 'C2': 'C4', 'C3': 'C2'}

r1

{'C3': 'r1', 'C1': 'r2', 'C4': 'r3', 'C2': 'r4'}

SVC

C1

Health Care Services Satisfaction Prediction

The prediction probability associated with class C2 and class C1, respectively, is 35.34% and 64.66%. Based on these probabilities, the model labels the given case as C1 since it is the most probable class. According to the attribution analysis, the most relevant features considered by the model here are F16, F7, and F10, while the least relevant features are F9, F14, and F12. Regarding the direction of influence of the features, F16, F7, F10, and F5 are the top positively supporting features, driving the decision higher in favour of C1. Further increasing the probability that C1 is the true label are the values of other positive features such as F1, F15, F13, and F6. To explain why the likelihood of C2 is 35.34%, we have to look at the negative contributions from F11, F8, F3, F14, F9, and F12. The abovementioned negative features contradict the model's decision with respect to the classification outcome.

{'C2': '35.34%', 'C1': '64.66%'}

{'F16': 'waiting rooms', 'F10': 'Hygiene and cleaning', 'F7': 'Specialists avaliable', 'F11': 'Quality\\/experience dr.', 'F5': 'Modern equipment', 'F8': 'Exact diagnosis', 'F1': 'hospital rooms quality', 'F13': 'Check up appointment', 'F3': 'avaliablity of drugs', 'F15': 'friendly health care workers', 'F6': 'Time waiting', 'F4': 'Communication with dr', 'F2': 'lab services', 'F14': 'parking, playing rooms, caffes', 'F9': 'Time of appointment', 'F12': 'Admin procedures'}

{'F14': 'F16', 'F4': 'F10', 'F7': 'F7', 'F6': 'F11', 'F10': 'F5', 'F9': 'F8', 'F15': 'F1', 'F1': 'F13', 'F13': 'F3', 'F11': 'F15', 'F2': 'F6', 'F8': 'F4', 'F12': 'F2', 'F16': 'F14', 'F5': 'F9', 'F3': 'F12'}

{'C2': 'C2', 'C1': 'C1'}

Satisfied

{'C2': 'Dissatisfied', 'C1': 'Satisfied'}

KNeighborsClassifier

C2

Real Estate Investment

The classifier is very uncertain about the correct label for the case given. Regarding the classifier's decision, there is close to an even split on the probability of either of the possible labels is the correct label but the classifier chooses the label as C2. The prediction verdict above is attributed to the contributions of mainly the following features: F14, F9, F12, and F4, however, the lowest ranked features are F18, F1, and F3. Analysing the direction of influence of the features shows that there are ten positive and ten negative features. Positive features such as F12, F4, F10, and F11 increase the response of the classifier in favour of the assigned label. Conversely, negative features such as F14, F9, F13, and F6 decrease the likelihood of C2 being the correct label given that their values support the alternative label, C1. The uncertainty concerning the label assignment can be due to the fact that the top negative features F14 and F9 have very high attributions shifting the classifier's verdict away from the C2 class.

{'C2': '50.00%', 'C1': '50.00%'}

{'F14': 'Feature7', 'F9': 'Feature4', 'F12': 'Feature2', 'F4': 'Feature8', 'F10': 'Feature20', 'F13': 'Feature1', 'F6': 'Feature12', 'F11': 'Feature15', 'F16': 'Feature6', 'F15': 'Feature9', 'F7': 'Feature17', 'F17': 'Feature3', 'F19': 'Feature19', 'F5': 'Feature13', 'F2': 'Feature18', 'F8': 'Feature5', 'F20': 'Feature11', 'F18': 'Feature16', 'F1': 'Feature10', 'F3': 'Feature14'}

{'F11': 'F14', 'F9': 'F9', 'F1': 'F12', 'F3': 'F4', 'F20': 'F10', 'F7': 'F13', 'F15': 'F6', 'F4': 'F11', 'F10': 'F16', 'F12': 'F15', 'F6': 'F7', 'F8': 'F17', 'F5': 'F19', 'F16': 'F5', 'F19': 'F2', 'F2': 'F8', 'F14': 'F20', 'F18': 'F18', 'F13': 'F1', 'F17': 'F3'}

{'C2': 'C2', 'C1': 'C1'}

Ignore

{'C2': 'Ignore', 'C1': 'Invest'}

DecisionTreeClassifier

C2

Car Acceptability Valuation

The classification algorithm believes that C2 is the output label that was generated with 100% certainty and that C1 is unlikely to be the correct label in this case. According to the attribution investigations, the following input features are ranked from most relevant to least relevant: F1, F5, F4, F2, F6, and F3. As shown by the attribution plot, F1 is the only one shown to positively contribute to the above classification decision, while the others contribute negatively. The contributions of negative features such as F5, F4, and F2 result in the decision being driven in a different direction. From the prediction confidence level, we can conclude that the very strong influence of F1 overshadows the contributions of the negative features hence the very high confidence level.

{'C2': '100.00%', 'C1': '0.00%'}

{'F1': 'safety', 'F5': 'persons', 'F4': 'buying', 'F2': 'maint', 'F6': 'lug_boot', 'F3': 'doors'}

{'F6': 'F1', 'F4': 'F5', 'F1': 'F4', 'F2': 'F2', 'F5': 'F6', 'F3': 'F3'}

{'C2': 'C2', 'C1': 'C1'}

Unacceptable

{'C2': 'Unacceptable', 'C1': 'Acceptable'}

RandomForestClassifier

C1

Flight Price-Range Classification

Of the three possible labels, there is 100.0% confidence that C1 is the most probable label for the given case. The features that heavily influence the classification verdict presented here are F6, F11, and F2, and they have a very strong positive contribution, increasing the odds of the C1 prediction. Other features with a positive influence on the model are F8, F1, F3, F7, and F10. On the contrary, F12, F9, and F4 make the model's decision fluctuate negatively towards selecting an alternative label. All of the negative features mentioned above have a low to moderate impact on the classification verdict presented here compared to F2, F11, and F6. Finally, F5 with its very low positive impact is the least ranked feature marginally pushing the decision towards the assigned label.

{'C1': '100.00%', 'C2': '0.00%', 'C3': '0.00%'}

{'F2': 'Duration_hours', 'F6': 'Airline', 'F11': 'Total_Stops', 'F1': 'Journey_day', 'F12': 'Source', 'F8': 'Destination', 'F10': 'Journey_month', 'F7': 'Dep_minute', 'F3': 'Arrival_minute', 'F9': 'Arrival_hour', 'F4': 'Duration_mins', 'F5': 'Dep_hour'}

{'F7': 'F2', 'F9': 'F6', 'F12': 'F11', 'F1': 'F1', 'F10': 'F12', 'F11': 'F8', 'F2': 'F10', 'F4': 'F7', 'F6': 'F3', 'F5': 'F9', 'F8': 'F4', 'F3': 'F5'}

{'C3': 'C1', 'C2': 'C2', 'C1': 'C3'}

Low

{'C1': 'Low', 'C2': 'Moderate', 'C3': 'High'}

RandomForestClassifier

C2

Used Cars Price-Range Prediction

Per the model, class C1 has a prediction probability of 10.50 percent, whereas class C2 has a predicted probability of 89.50 percent. As a result of the model, it can be determined that C2 is the most likely label for the given scenario. All of the input features are shown to contribute to the above conclusion, with F1, F7, and F6 having the most influence on the classification decision. The least influential features with regard to this classification are F5, F3, F2, and F9, whereas, the impact of F8, F4, and F10 can be classified as modest. The large positive contributions of F7 and F1 are responsible for the model's high confidence which further supported by the positive contributions of F8, F5, and F3. In conclusion, the negative features F6, F4, F2, F10, and F9 favour labelling the case as C1 hence the associated predicted probability.

{'C1': '10.50%', 'C2': '89.50%'}

{'F7': 'Power', 'F1': 'car_age', 'F6': 'Transmission', 'F8': 'Fuel_Type', 'F4': 'Name', 'F10': 'Mileage', 'F5': 'Engine', 'F2': 'Owner_Type', 'F3': 'Kilometers_Driven', 'F9': 'Seats'}

{'F4': 'F7', 'F5': 'F1', 'F8': 'F6', 'F7': 'F8', 'F6': 'F4', 'F2': 'F10', 'F3': 'F5', 'F9': 'F2', 'F1': 'F3', 'F10': 'F9'}

{'C2': 'C1', 'C1': 'C2'}

High

{'C1': 'Low', 'C2': 'High'}

AdaBoostClassifier

C1

Basketball Players Career Length Prediction

The classifier says that C1 is the most likely label for the provided data with relatively high confidence. It is crucial to remember, however, that there is a 21.80% possibility that it is C2. F7 and F19 are the major driving variables for the aforementioned classification or prediction choice. The remaining variables F9, F11, F1, and F3 have a modest to minor impact on the selection made above. Among the input variables, F9, F3, F4, F15, and F12 are the subset that have a negative influence or contribution whereas all of the remaining variables have a positive impact. In essence, the substantial positive contributions of F7 and F19, together with the contributions of additional positive variables such as F11, F1, F6, and F13, account for the classifier's confidence in this classification.

{'C1': '78.20%', 'C2': '21.80%'}

{'F7': 'GamesPlayed', 'F19': 'PointsPerGame', 'F9': 'Steals', 'F11': 'MinutesPlayed', 'F1': 'DefensiveRebounds', 'F3': 'Rebounds', 'F6': 'Blocks', 'F13': 'FreeThrowAttempt', 'F4': 'FieldGoalPercent', 'F18': 'FreeThrowMade', 'F15': 'OffensiveRebounds', 'F17': 'FieldGoalsMade', 'F2': '3PointAttempt', 'F10': 'FreeThrowPercent', 'F16': '3PointMade', 'F5': 'FieldGoalsAttempt', 'F8': 'Turnovers', 'F12': 'Assists', 'F14': '3PointPercent'}

{'F1': 'F7', 'F3': 'F19', 'F17': 'F9', 'F2': 'F11', 'F14': 'F1', 'F15': 'F3', 'F18': 'F6', 'F11': 'F13', 'F6': 'F4', 'F10': 'F18', 'F13': 'F15', 'F4': 'F17', 'F8': 'F2', 'F12': 'F10', 'F7': 'F16', 'F5': 'F5', 'F19': 'F8', 'F16': 'F12', 'F9': 'F14'}

{'C2': 'C1', 'C1': 'C2'}

More than 5

{'C1': 'More than 5', 'C2': 'Less than 5'}

RandomForestClassifier

C1

Printer Sales

C1 has an 83.0% chance of being the correct label for the case under consideration, making C2 the least likely class with a predicted likelihood of 17.0%. F24, F2, and F21 features have a significant impact on class selection here while on the other hand, the remaining features are shown to have marginal to no contribution to the classification verdict here. In actual fact, the values for F7, F5, F1, F6, F11, and F3 may have been ignored by the classifier because their respective influences are almost zero. Of the important features, only F12, F23, F13, F14, F20, and F9 are negative and this is mainly because their contribution to selection tends to reduce the chance that C1 is the correct label, preferring that the case is classified as C2. The remaining features such as F24, F2, F21, F17, F15, and F22 strongly contribute positively, increasing the chances of C1 which explains the level of certainty associated with C1.

{'C1': '83.00%', 'C2': '17.00%'}

{'F24': 'X8', 'F2': 'X24', 'F21': 'X1', 'F15': 'X2', 'F17': 'X10', 'F22': 'X15', 'F12': 'X25', 'F8': 'X23', 'F4': 'X18', 'F23': 'X4', 'F13': 'X7', 'F19': 'X17', 'F16': 'X3', 'F26': 'X22', 'F14': 'X5', 'F20': 'X9', 'F9': 'X12', 'F10': 'X19', 'F25': 'X11', 'F18': 'X16', 'F7': 'X14', 'F5': 'X21', 'F1': 'X20', 'F6': 'X13', 'F11': 'X6', 'F3': 'X26'}

{'F8': 'F24', 'F24': 'F2', 'F1': 'F21', 'F2': 'F15', 'F10': 'F17', 'F15': 'F22', 'F25': 'F12', 'F23': 'F8', 'F18': 'F4', 'F4': 'F23', 'F7': 'F13', 'F17': 'F19', 'F3': 'F16', 'F22': 'F26', 'F5': 'F14', 'F9': 'F20', 'F12': 'F9', 'F19': 'F10', 'F11': 'F25', 'F16': 'F18', 'F14': 'F7', 'F21': 'F5', 'F20': 'F1', 'F13': 'F6', 'F6': 'F11', 'F26': 'F3'}

{'C1': 'C1', 'C2': 'C2'}

Less

{'C1': 'Less', 'C2': 'More'}

LogisticRegression

C1

Music Concert Attendance

C1 is the label picked by the algorithm with about 82.06% certainty, since the prediction likelihood of C2 is only 17.94%. F6, F12, F13, and F14 all contribute significantly to the above classification output and among them, the features that support the most positive contribution to the C1 prediction are F14, F6, and F12, while F13 drives the final prediction against assigning C1 in support of C2. F16 also contributes positively to the classification here, but F1 contributes negatively and like F13 favours C2. Finally, according to the analysis, F19, F18, F3, and F11 all have little effect on the final prediction made by the algorithm for this case.

{'C2': '17.94%', 'C1': '82.06%'}

{'F14': 'X11', 'F6': 'X1', 'F13': 'X13', 'F12': 'X3', 'F1': 'X8', 'F16': 'X6', 'F9': 'X2', 'F20': 'X9', 'F2': 'X17', 'F15': 'X10', 'F5': 'X4', 'F10': 'X14', 'F17': 'X20', 'F7': 'X18', 'F4': 'X19', 'F8': 'X7', 'F3': 'X12', 'F19': 'X15', 'F11': 'X16', 'F18': 'X5'}

{'F11': 'F14', 'F1': 'F6', 'F13': 'F13', 'F3': 'F12', 'F8': 'F1', 'F6': 'F16', 'F2': 'F9', 'F9': 'F20', 'F17': 'F2', 'F10': 'F15', 'F4': 'F5', 'F14': 'F10', 'F20': 'F17', 'F18': 'F7', 'F19': 'F4', 'F7': 'F8', 'F12': 'F3', 'F15': 'F19', 'F16': 'F11', 'F5': 'F18'}

{'C1': 'C2', 'C2': 'C1'}

> 10k

{'C2': '< 10k', 'C1': '> 10k'}

LogisticRegression

C3

Flight Price-Range Classification

The model is confident in its prediction, as it predicted class C3 with a likelihood of 90.48% and hence, for the given case, there is a smaller chance of it being any other class label. F7 and F1 are deemed the most important features whereas on the other hand all the other features have moderate to minimal amounts of influence. Both F7 and F1 have the same direction of impact, increasing the odds of the predicted label, C3. While F2 and F11 are both encouraging the model to make a prediction of C3, the others F3, F9, and F8 is pushing the model towards a different label. Many features have moderately low impact on the final prediction, but the features F10, F8, and F4 are those with the smallest influence.

{'C3': '90.48%', 'C2': '9.51%', 'C1': '0.01%'}

{'F7': 'Total_Stops', 'F1': 'Airline', 'F2': 'Destination', 'F11': 'Arrival_hour', 'F3': 'Source', 'F5': 'Duration_hours', 'F6': 'Dep_hour', 'F9': 'Dep_minute', 'F12': 'Arrival_minute', 'F10': 'Journey_month', 'F8': 'Journey_day', 'F4': 'Duration_mins'}

{'F12': 'F7', 'F9': 'F1', 'F11': 'F2', 'F5': 'F11', 'F10': 'F3', 'F7': 'F5', 'F3': 'F6', 'F4': 'F9', 'F6': 'F12', 'F2': 'F10', 'F1': 'F8', 'F8': 'F4'}

{'C1': 'C3', 'C3': 'C2', 'C2': 'C1'}

Low

{'C3': 'Low', 'C2': 'Moderate', 'C1': 'High'}

DecisionTreeClassifier

C1

Airline Passenger Satisfaction

Based on the probability distribution across the classes, the classifier is shown to have a moderately high confidence level in the C1 label assignment, with its likelihood equal to 65.0%, whereas that of C2 is only 35.0%. The prediction decision above is predominantly due to the influence of the variables F19, F18, F17, and F8. On the lower end are the least relevant variables, F14, F22, F25, F16, F24, and F6, with little to no influence on the classifier when assigning a label to the given instance. On the one hand, the top positive variables are F19, F18, and F17, increasing the probability that C1 is the correct label. Also, the top negative variables are F8, F2, F9, and F5, decreasing the classifier's response and consequently shifting the prediction verdict in the opposite direction towards C2. Other variables with a positive direction of influence are F21, F13, F20, F7, F1, F12, F26, F4, and F3.

{'C2': '35.00%', 'C1': '65.00%'}

{'F19': 'X8', 'F18': 'X2', 'F17': 'X1', 'F8': 'X21', 'F21': 'X25', 'F2': 'X10', 'F9': 'X3', 'F5': 'X9', 'F13': 'X15', 'F10': 'X7', 'F23': 'X20', 'F15': 'X12', 'F20': 'X24', 'F7': 'X6', 'F1': 'X17', 'F12': 'X23', 'F11': 'X11', 'F26': 'X22', 'F4': 'X4', 'F3': 'X14', 'F14': 'X19', 'F22': 'X18', 'F25': 'X16', 'F16': 'X13', 'F24': 'X5', 'F6': 'X26'}

{'F8': 'F19', 'F2': 'F18', 'F1': 'F17', 'F21': 'F8', 'F25': 'F21', 'F10': 'F2', 'F3': 'F9', 'F9': 'F5', 'F15': 'F13', 'F7': 'F10', 'F20': 'F23', 'F12': 'F15', 'F24': 'F20', 'F6': 'F7', 'F17': 'F1', 'F23': 'F12', 'F11': 'F11', 'F22': 'F26', 'F4': 'F4', 'F14': 'F3', 'F19': 'F14', 'F18': 'F22', 'F16': 'F25', 'F13': 'F16', 'F5': 'F24', 'F26': 'F6'}

{'C1': 'C2', 'C2': 'C1'}

Acceptable

{'C2': 'neutral or dissatisfied', 'C1': 'satisfied'}

KNeighborsClassifier

C2

Company Bankruptcy Prediction

The model's output labelling judgement for the case under consideration is as follows: C1 cannot be the label for the given case; C2 is the most likely class label with a 100.0% confidence level. The key driving factors resulting in the aforementioned classification are the values of the input features: F9, F46, F8, F88, F42, F38, and F92. F61, F69, F19, F36, F89, F17, F15, F30, F26, F27, F7, F62, and F3 are the features that have a modest effect on the decision. Aside from the aforementioned input features, all others, such as F50, F34, F18, and F49, are revealed to be irrelevant to the conclusion reached here. Not all of the influential features support labelling the current instance as C2, and they are referred to as negative features. F92, F69, F7, F62, and F3 are the negative attributes that diminish the likelihood that C2 is the correct label in this case. F9, F46, F8, and F88 are important positive features that strongly increase the likelihood that C2 is the correct label.

{'C2': '100.00%', 'C1': '0.00%'}

{'F9': ' Interest Coverage Ratio (Interest expense to EBIT)', 'F46': ' Net Income to Total Assets', 'F8': ' Realized Sales Gross Profit Growth Rate', 'F88': ' Accounts Receivable Turnover', 'F42': ' Operating Expense Rate', 'F92': ' Contingent liabilities\\/Net worth', 'F38': ' Non-industry income and expenditure\\/revenue', 'F61': ' Current Ratio', 'F69': ' Cash Flow to Liability', 'F19': ' Fixed Assets Turnover Frequency', 'F89': ' Regular Net Profit Growth Rate', 'F36': ' Quick Asset Turnover Rate', 'F17': ' Net Value Per Share (C)', 'F15': ' Operating Profit Growth Rate', 'F30': ' After-tax Net Profit Growth Rate', 'F26': ' Continuous Net Profit Growth Rate', 'F27': ' Net Value Per Share (B)', 'F7': ' Equity to Long-term Liability', 'F62': ' CFO to Assets', 'F3': ' Total debt\\/Total net worth', 'F50': ' Current Asset Turnover Rate', 'F34': " Net Income to Stockholder's Equity", 'F18': ' Operating Gross Margin', 'F49': ' Operating Profit Per Share (Yuan ¥)', 'F55': ' Operating Profit Rate', 'F91': ' Cash Flow Per Share', 'F57': ' Total income\\/Total expense', 'F37': ' No-credit Interval', 'F5': ' Liability to Equity', 'F87': ' Working Capital to Total Assets', 'F43': ' Working Capital\\/Equity', 'F56': ' Long-term Liability to Current Assets', 'F2': ' Interest-bearing debt interest rate', 'F45': ' Inventory and accounts receivable\\/Net value', 'F41': ' Realized Sales Gross Margin', 'F85': ' Current Liability to Equity', 'F90': ' Equity to Liability', 'F71': ' Current Liability to Liability', 'F21': ' Operating profit\\/Paid-in capital', 'F93': ' Operating Funds to Liability', 'F82': ' Current Liability to Current Assets', 'F16': ' Net worth\\/Assets', 'F59': ' Tax rate (A)', 'F48': ' Quick Assets\\/Current Liability', 'F31': ' After-tax net Interest Rate', 'F84': ' Per Share Net profit before tax (Yuan ¥)', 'F32': ' Total Asset Turnover', 'F63': ' Cash Reinvestment %', 'F1': ' Fixed Assets to Assets', 'F14': ' Working capitcal Turnover Rate', 'F64': ' Net profit before tax\\/Paid-in capital', 'F78': ' Net Worth Turnover Rate (times)', 'F20': ' Debt ratio %', 'F29': ' Cash Flow to Equity', 'F44': ' Long-term fund suitability ratio (A)', 'F6': ' Cash Flow to Sales', 'F66': ' Total Asset Growth Rate', 'F68': ' Inventory\\/Current Liability', 'F70': ' Allocation rate per person', 'F22': ' Inventory Turnover Rate (times)', 'F83': ' Operating profit per person', 'F10': ' Net Value Growth Rate', 'F80': ' Interest Expense Ratio', 'F52': ' ROA(B) before interest and depreciation after tax', 'F24': ' Continuous interest rate (after tax)', 'F76': ' Inventory\\/Working Capital', 'F4': ' Retained Earnings to Total Assets', 'F86': ' Total assets to GNP price', 'F73': ' Persistent EPS in the Last Four Seasons', 'F67': ' Quick Ratio', 'F11': ' Revenue per person', 'F58': ' Borrowing dependency', 'F13': ' Cash\\/Total Assets', 'F60': ' ROA(A) before interest and % after tax', 'F75': ' ROA(C) before interest and depreciation before interest', 'F65': ' Average Collection Days', 'F53': ' Current Liabilities\\/Liability', 'F33': ' Cash Flow to Total Assets', 'F28': ' Pre-tax net Interest Rate', 'F35': ' Current Liability to Assets', 'F40': ' Quick Assets\\/Total Assets', 'F12': ' Total expense\\/Assets', 'F51': ' Net Value Per Share (A)', 'F81': ' Current Assets\\/Total Assets', 'F39': ' Research and development expense rate', 'F74': ' Current Liabilities\\/Equity', 'F72': ' Cash flow rate', 'F77': ' Total Asset Return Growth Rate Ratio', 'F25': ' Degree of Financial Leverage (DFL)', 'F23': ' Cash Turnover Rate', 'F47': ' Cash\\/Current Liability', 'F79': ' Revenue Per Share (Yuan ¥)', 'F54': ' Gross Profit to Sales'}

{'F60': 'F9', 'F16': 'F46', 'F38': 'F8', 'F2': 'F88', 'F19': 'F42', 'F64': 'F92', 'F4': 'F38', 'F82': 'F61', 'F50': 'F69', 'F22': 'F19', 'F85': 'F89', 'F33': 'F36', 'F88': 'F17', 'F43': 'F15', 'F80': 'F30', 'F54': 'F26', 'F27': 'F27', 'F23': 'F7', 'F76': 'F62', 'F7': 'F3', 'F61': 'F50', 'F59': 'F34', 'F62': 'F18', 'F63': 'F49', 'F58': 'F55', 'F65': 'F91', 'F57': 'F57', 'F56': 'F37', 'F66': 'F5', 'F67': 'F87', 'F68': 'F43', 'F69': 'F56', 'F1': 'F2', 'F70': 'F45', 'F83': 'F41', 'F92': 'F85', 'F91': 'F90', 'F90': 'F71', 'F89': 'F21', 'F87': 'F93', 'F86': 'F82', 'F84': 'F16', 'F81': 'F59', 'F71': 'F48', 'F79': 'F31', 'F78': 'F84', 'F77': 'F32', 'F75': 'F63', 'F74': 'F1', 'F73': 'F14', 'F72': 'F64', 'F55': 'F78', 'F47': 'F20', 'F53': 'F29', 'F52': 'F44', 'F25': 'F6', 'F24': 'F66', 'F21': 'F68', 'F20': 'F70', 'F18': 'F22', 'F17': 'F83', 'F15': 'F10', 'F14': 'F80', 'F13': 'F52', 'F12': 'F24', 'F11': 'F76', 'F10': 'F4', 'F9': 'F86', 'F8': 'F73', 'F6': 'F67', 'F5': 'F11', 'F3': 'F58', 'F26': 'F13', 'F28': 'F60', 'F29': 'F75', 'F41': 'F65', 'F51': 'F53', 'F49': 'F33', 'F48': 'F28', 'F46': 'F35', 'F45': 'F40', 'F44': 'F12', 'F42': 'F51', 'F40': 'F81', 'F30': 'F39', 'F39': 'F74', 'F37': 'F72', 'F36': 'F77', 'F35': 'F25', 'F34': 'F23', 'F32': 'F47', 'F31': 'F79', 'F93': 'F54'}

{'C1': 'C2', 'C2': 'C1'}

No

{'C2': 'No', 'C1': 'Yes'}

BernoulliNB

C2

Student Job Placement

For the case under consideration, the model assigned C2 with very high confidence, since the likelihood of C1 being the right label is only 0.52% which is very small. F9, F3, F1, and F7 have a large positive impact on the model's output prediction. F1 and F7 have a moderately positive impact on the prediction of C2, while F12 has a similar impact but in the opposite direction. F8, F4, and F5 have a very low impact on classification. F11, F2, F10, and F6 have a larger but still insignificant effect. Examining the attributions indicates that there are only two features, F12 and F4, with values that contradict the prediction made here but, their impact on the model is smaller when compared to positive features such as F3, F1, and F9, which explains why the confidence level associated with this classification is high.

{'C1': '0.52%', 'C2': '99.48%'}

{'F9': 'workex', 'F3': 'specialisation', 'F1': 'ssc_p', 'F7': 'hsc_p', 'F12': 'degree_p', 'F11': 'gender', 'F6': 'degree_t', 'F10': 'etest_p', 'F2': 'hsc_b', 'F4': 'hsc_s', 'F5': 'ssc_b', 'F8': 'mba_p'}

{'F11': 'F9', 'F12': 'F3', 'F1': 'F1', 'F2': 'F7', 'F3': 'F12', 'F6': 'F11', 'F10': 'F6', 'F4': 'F10', 'F8': 'F2', 'F9': 'F4', 'F7': 'F5', 'F5': 'F8'}

{'C2': 'C1', 'C1': 'C2'}

Placed

{'C1': 'Not Placed', 'C2': 'Placed'}

LogisticRegression

C3

Flight Price-Range Classification

Since the likelihood of C3 being the true label is shown by the prediction algorithm outputs to be equal to 93.02 percent, there is only a small chance that the true label for the given data instance is any of the other class labels, C2 and C1. The features F2, F8, F9, and F12 are the most important ones driving the label assignment verdict above, and on the other hand, the least relevant features are shown to be F1, F10, and F6. Considering the direction of influence of each input feature, as shown by the attribution analysis, it can be concluded that the positive features steering the prediction higher towards C3 are F2, F8, F12, F9, F4, F11, and F10. The marginal doubt in the predicted output decision is attributed to the negative contributions of F5, F3, F6, F1, and F7. Considering the attributions of the features and predicted probabilities across the classes, it can be concluded that the joint positive contribution outranks the negative contributions; hence, the algorithm is confident that C3 is likely the true label.

{'C3': '93.02%', 'C2': '6.97%', 'C1': '0.01%'}

{'F2': 'Total_Stops', 'F8': 'Airline', 'F12': 'Destination', 'F9': 'Journey_day', 'F5': 'Source', 'F3': 'Dep_hour', 'F4': 'Duration_hours', 'F7': 'Dep_minute', 'F11': 'Duration_mins', 'F1': 'Arrival_minute', 'F10': 'Arrival_hour', 'F6': 'Journey_month'}

{'F12': 'F2', 'F9': 'F8', 'F11': 'F12', 'F1': 'F9', 'F10': 'F5', 'F3': 'F3', 'F7': 'F4', 'F4': 'F7', 'F8': 'F11', 'F6': 'F1', 'F5': 'F10', 'F2': 'F6'}

{'C3': 'C3', 'C2': 'C2', 'C1': 'C1'}

Low

{'C3': 'Low', 'C2': 'Moderate', 'C1': 'High'}

RandomForestClassifier

C1

Cab Surge Pricing System

Between the three possible classes, there is an 88.0% probability that the correct label for this case is C1. This means that there is a 12.0% chance that the label could be one of the other possible labels, C2 or C3. Increasing the odds of the predicted label are the variables F3, F1, F12, and F2. The next set of variables, F5, F8, and F11, have values that moderately decrease the likelihood of C1 being the correct label. F7, F9, and F6 are the other negatively contributing features, and given that they are lowly ranked, they have a marginal impact when determining the correct label for this case. The other positive features further increasing the probability that C1 is the right label are F4 and F10. Overall, we can conclude that the decision to label the case as C1 is largely due to the strong positive influence of F1, F3, F2, and F12.

{'C2': '3.00%', 'C3': '9.00%', 'C1': '88.00%'}

{'F3': 'Type_of_Cab', 'F1': 'Destination_Type', 'F12': 'Cancellation_Last_1Month', 'F2': 'Trip_Distance', 'F5': 'Customer_Rating', 'F8': 'Life_Style_Index', 'F11': 'Var3', 'F4': 'Var1', 'F10': 'Customer_Since_Months', 'F7': 'Var2', 'F9': 'Gender', 'F6': 'Confidence_Life_Style_Index'}

{'F2': 'F3', 'F6': 'F1', 'F8': 'F12', 'F1': 'F2', 'F7': 'F5', 'F4': 'F8', 'F11': 'F11', 'F9': 'F4', 'F3': 'F10', 'F10': 'F7', 'F12': 'F9', 'F5': 'F6'}

{'C2': 'C2', 'C3': 'C3', 'C1': 'C1'}

C3

{'C2': 'Low', 'C3': 'Medium', 'C1': 'High'}

RandomForestClassifier

C2

Wine Quality Prediction

Based on the input variables, the model is moderately confident that the C2 is the appropriate label for the data under consideration. As a matter of fact, the prediction likelihood associated with class C1 is about 30.42%. The preceeding classification verdict can be largely blamed on the contributions of variables F11, F5, F3, and F9, whereas those with marginally lower contributions are F2, F8, and F1. The variables with moderate contributions are F6, F4, F7, and F10. Considering their respective contributions, F11, F3, F9, and F10 are the variables with positive influence that increase the chances of C2 being the correct label for the given data. The little doubt in the label choice here could be attributed to the negative variables, mainly F5, F6, F7, and F4, which decrease the chances of the model labelling the data given as C2 since these negative variables favour selecting the alternative label, C1 over C2. Given that majority of top variables contribute positively, it is not unexpected that C2 is the picked label with reasonably high confidence.

{'C1': '30.42%', 'C2': '69.58%'}

{'F11': 'alcohol', 'F5': 'sulphates', 'F3': 'volatile acidity', 'F9': 'total sulfur dioxide', 'F6': 'fixed acidity', 'F4': 'citric acid', 'F7': 'residual sugar', 'F10': 'density', 'F2': 'chlorides', 'F8': 'pH', 'F1': 'free sulfur dioxide'}

{'F11': 'F11', 'F10': 'F5', 'F2': 'F3', 'F7': 'F9', 'F1': 'F6', 'F3': 'F4', 'F4': 'F7', 'F8': 'F10', 'F5': 'F2', 'F9': 'F8', 'F6': 'F1'}

{'C1': 'C1', 'C2': 'C2'}

high quality

{'C1': 'low_quality', 'C2': 'high quality'}

SVC

C1

E-Commerce Shipping

The classifier is 69.02% certain that the given case is under the class label C1, implying that the likelihood of C2 is only 30.98%. Analysis performed to understand the contribution of each input feature revealed that: F9, F4, and F8 are the most influential features when assigning a label to the given case. Features F1, F6, F5, and F2 have moderate contributions, whereas the F10, F3 and F7 have lower relevance to the final classification decision. F9 and F8 push the class assignment towards C1, whereas F4 does the opposite, decreasing the likelihood of C1. Similar to F4, F1, and F6 negatively impact the C1 classification, whereas F2, F10, and F5 positively push the decision towards the C1 class. Features F3, and F7 all have little impact on the final decision, with F7 having the least impact.

{'C1': '69.02%', 'C2': '30.98%'}

{'F9': 'Weight_in_gms', 'F4': 'Discount_offered', 'F8': 'Prior_purchases', 'F1': 'Customer_care_calls', 'F6': 'Cost_of_the_Product', 'F5': 'Mode_of_Shipment', 'F2': 'Customer_rating', 'F10': 'Gender', 'F3': 'Product_importance', 'F7': 'Warehouse_block'}

{'F3': 'F9', 'F2': 'F4', 'F8': 'F8', 'F6': 'F1', 'F1': 'F6', 'F5': 'F5', 'F7': 'F2', 'F10': 'F10', 'F9': 'F3', 'F4': 'F7'}

{'C1': 'C1', 'C2': 'C2'}

On-time

{'C1': 'On-time', 'C2': 'Late'}

RandomForestClassifier

C1

Advertisement Prediction

The classifier trained on this prediction problem assigns a label to a given case based on the information supplied. The class assigned by the classifier to the case under consideration is C1. The probability that C2 is the correct label is around 25.28%; therefore, it is less likely to be the true label. The above classification decision is mainly based on the influence of the features F5, F3, F2, F1, F4, F7, and F6. Of the above stated features, F1 and F3 are the ones shown to have a negative impact, decreasing the odds of C1 being the accurate label for the given case and encouraging the classifier to select C2 instead. Finally, it can be concluded that there is a moderately high level of confidence in the assigned label, which can be attributed to the strong positive contribution of F5 combined with other positive features such as F2 and F4.

{'C1': '74.72%', 'C2': '25.28%'}

{'F5': 'Daily Time Spent on Site', 'F3': 'Daily Internet Usage', 'F2': 'Age', 'F1': 'ad_day', 'F4': 'Area Income', 'F7': 'Gender', 'F6': 'ad_month'}

{'F1': 'F5', 'F4': 'F3', 'F2': 'F2', 'F7': 'F1', 'F3': 'F4', 'F5': 'F7', 'F6': 'F6'}

{'C1': 'C1', 'C2': 'C2'}

Skip

{'C1': 'Skip', 'C2': 'Watch'}

GradientBoostingClassifier

C1

Food Ordering Customer Churn Prediction

The case given is labelled as C1 by the classifier with a confidence level equal to 82.07%. Therefore, the probability of C2 being the correct label is only 17.93%. The classification above is mainly due to the contributions of features such as F29, F13, F46, and F28. F1, F40, and F11 are the next three with moderate influence. However, not all the features are considered by the classifier when determining the correct label for the given case. F27, F21, F15, and F25 are notable irrelevant features. With regards to the direction of influence of the relevant features, F29, F13, F46, and F28 are the top features with strong positive contributions favouring the assignment of label C1. The top negative features that shift the classification in a different direction are F1, F40, F20, and F36. Considering the fact that a number of the relevant features have positive attributions, it is not surprising that the classifier is quite certain that the appropriate label is C1 instead of C2.

{'C2': '17.93%', 'C1': '82.07%'}

{'F29': 'More restaurant choices', 'F13': 'Ease and convenient', 'F46': 'Bad past experience', 'F28': 'Time saving', 'F1': 'Easy Payment option', 'F40': 'Good Tracking system', 'F11': 'Wrong order delivered', 'F20': 'Influence of rating', 'F12': 'Late Delivery', 'F36': 'Less Delivery time', 'F8': 'Long delivery time', 'F10': 'Delivery person ability', 'F38': 'Order placed by mistake', 'F30': 'More Offers and Discount', 'F7': 'Freshness ', 'F35': 'Unavailability', 'F32': 'Delay of delivery person picking up food', 'F45': 'Poor Hygiene', 'F37': 'Order Time', 'F6': 'Delay of delivery person getting assigned', 'F27': 'High Quality of package', 'F21': 'Residence in busy location', 'F15': 'Good Taste ', 'F25': 'Temperature', 'F9': 'Google Maps Accuracy', 'F18': 'Good Road Condition', 'F16': 'Number of calls', 'F34': 'Low quantity low time', 'F2': 'Politeness', 'F26': 'Maximum wait time', 'F39': 'Age', 'F43': 'Influence of time', 'F19': 'Missing item', 'F44': 'Family size', 'F42': 'Unaffordable', 'F31': 'Health Concern', 'F41': 'Self Cooking', 'F24': 'Good Food quality', 'F17': 'Perference(P2)', 'F33': 'Perference(P1)', 'F3': 'Educational Qualifications', 'F23': 'Monthly Income', 'F4': 'Occupation', 'F22': 'Marital Status', 'F5': 'Gender', 'F14': 'Good Quantity'}

{'F12': 'F29', 'F10': 'F13', 'F21': 'F46', 'F11': 'F28', 'F13': 'F1', 'F16': 'F40', 'F27': 'F11', 'F38': 'F20', 'F19': 'F12', 'F39': 'F36', 'F24': 'F8', 'F37': 'F10', 'F29': 'F38', 'F14': 'F30', 'F43': 'F7', 'F22': 'F35', 'F26': 'F32', 'F20': 'F45', 'F31': 'F37', 'F25': 'F6', 'F40': 'F27', 'F33': 'F21', 'F45': 'F15', 'F44': 'F25', 'F34': 'F9', 'F35': 'F18', 'F41': 'F16', 'F36': 'F34', 'F42': 'F2', 'F32': 'F26', 'F1': 'F39', 'F30': 'F43', 'F28': 'F19', 'F7': 'F44', 'F23': 'F42', 'F18': 'F31', 'F17': 'F41', 'F15': 'F24', 'F9': 'F17', 'F8': 'F33', 'F6': 'F3', 'F5': 'F23', 'F4': 'F4', 'F3': 'F22', 'F2': 'F5', 'F46': 'F14'}

{'C2': 'C2', 'C1': 'C1'}

Go Away

{'C2': 'Return', 'C1': 'Go Away'}

SVM_linear

C3

Mobile Price-Range Classification

According to the algorithm, there is little to no chance that the correct label for the given data instance is any of the following classes: C4, C1, and C2. It is very confident that the proper label is C3. This label assignment is largely due to the parts played by the features F3, F18, and F2. On the lower end are the input features F15, F14, F7, and F13, which are shown to be less relevant when it comes to this labelling assignment task. Finally, among the top features identified during the attribution investogation, only F9 and F1 are features with a negative influence, decreasing the odds of C3 being the appropriate label here.

{'C4': '0.00%', 'C1': '0.00%', 'C2': '0.00%', 'C3': '100.00%'}

{'F3': 'ram', 'F18': 'battery_power', 'F2': 'px_width', 'F1': 'int_memory', 'F9': 'sc_h', 'F12': 'pc', 'F6': 'mobile_wt', 'F19': 'fc', 'F11': 'n_cores', 'F20': 'clock_speed', 'F5': 'blue', 'F8': 'three_g', 'F4': 'touch_screen', 'F16': 'm_dep', 'F17': 'px_height', 'F10': 'talk_time', 'F15': 'dual_sim', 'F14': 'wifi', 'F7': 'four_g', 'F13': 'sc_w'}

{'F11': 'F3', 'F1': 'F18', 'F10': 'F2', 'F4': 'F1', 'F12': 'F9', 'F8': 'F12', 'F6': 'F6', 'F3': 'F19', 'F7': 'F11', 'F2': 'F20', 'F15': 'F5', 'F18': 'F8', 'F19': 'F4', 'F5': 'F16', 'F9': 'F17', 'F14': 'F10', 'F16': 'F15', 'F20': 'F14', 'F17': 'F7', 'F13': 'F13'}

{'C4': 'C4', 'C1': 'C1', 'C3': 'C2', 'C2': 'C3'}

r4

{'C4': 'r1', 'C1': 'r2', 'C2': 'r3', 'C3': 'r4'}

SVC

C1

Paris House Classification

The model predicts that the label for this case is C1 with a high degree of certainty of about 99.19% and the probability of the other label is only 0.81%. From the analysis, the variables with the strongest attributions to this classification decision are F9, F14, and F13. The attributions of these variables increased the response of the model in favour of labelling the case as C1. Other variables that positively supported the label decision include F10, F7, and F8. Not all the variables support the model's prediction of C1 and this is because the values of F12, F6, F3, F5, and F1 are driving the prediction towards C2. The joint attribution from these variables is weaker than that from F9, F14, and F13, so the model is biased toward predicting C1. Finally, F15, F11, F17, and F2 are the least important positive features, given that they have minimal attributions in favour of C1.

{'C1': '99.19%', 'C2': '0.81%'}

{'F9': 'isNewBuilt', 'F13': 'hasYard', 'F14': 'hasPool', 'F12': 'hasStormProtector', 'F6': 'hasStorageRoom', 'F10': 'made', 'F7': 'basement', 'F8': 'numberOfRooms', 'F3': 'squareMeters', 'F4': 'floors', 'F5': 'numPrevOwners', 'F16': 'garage', 'F1': 'attic', 'F15': 'cityCode', 'F11': 'price', 'F17': 'cityPartRange', 'F2': 'hasGuestRoom'}

{'F3': 'F9', 'F1': 'F13', 'F2': 'F14', 'F4': 'F12', 'F5': 'F6', 'F12': 'F10', 'F13': 'F7', 'F7': 'F8', 'F6': 'F3', 'F8': 'F4', 'F11': 'F5', 'F15': 'F16', 'F14': 'F1', 'F9': 'F15', 'F17': 'F11', 'F10': 'F17', 'F16': 'F2'}

{'C2': 'C1', 'C1': 'C2'}

Basic

{'C1': 'Basic', 'C2': 'Luxury'}

LogisticRegression

C1

Used Cars Price-Range Prediction

According to the output prediction probabilities across the two classes, the output decision for the given data is C1 with a very high confidence level. C2 has a prediction probability of about 0.00%. The variables contributing most to the abovementioned classification are F1, F10, and F2, whereas F4 and F7 are the least influential variables. The very high confidence level associated with the classification decision here could be attributed to the fact that a greater number of the input variables have attributions that increase the model's response towards label C1. F6, F5, and F4 are the variables with negative contributions that attempt to push the model to label this case as C2. To put it in a nutshell, the joint contribution of the negative variables is very low unlike that of the positive variables, hence the model's certainty in the decision here.

{'C1': '100.00%', 'C2': '0.00%'}

{'F1': 'car_age', 'F10': 'Power', 'F2': 'Fuel_Type', 'F9': 'Engine', 'F8': 'Seats', 'F3': 'Transmission', 'F6': 'Kilometers_Driven', 'F5': 'Name', 'F4': 'Mileage', 'F7': 'Owner_Type'}

{'F5': 'F1', 'F4': 'F10', 'F7': 'F2', 'F3': 'F9', 'F10': 'F8', 'F8': 'F3', 'F1': 'F6', 'F6': 'F5', 'F2': 'F4', 'F9': 'F7'}

{'C2': 'C1', 'C1': 'C2'}

Low

{'C1': 'Low', 'C2': 'High'}

SVC

C2

Tic-Tac-Toe Strategy

With a labelling confidence level of 99.50%, the classifier predicts the label C2 in this situation. Hence, it is correct to conclude that the classifier is less certain that C1 is the proper label for the case here. The analysis indicates that five features contradict the decision above, while four features support the classifier. The features contradicting the prediction are usually referred to as negative features while those supporting it are referred to as positive features. The negative features decreasing the odds of C2 being the correct label are F7, F9, F5, F2, and F1. Conversely, the positive features increasing the odds of C2 are F4, F6, F8, and F3.

{'C1': '0.50%', 'C2': '99.50%'}

{'F7': 'middle-middle-square', 'F4': 'top-left-square', 'F6': 'bottom-left-square', 'F8': 'bottom-right-square', 'F9': ' top-right-square', 'F5': 'middle-right-square', 'F2': 'top-middle-square', 'F3': 'middle-left-square', 'F1': 'bottom-middle-square'}

{'F5': 'F7', 'F1': 'F4', 'F7': 'F6', 'F9': 'F8', 'F3': 'F9', 'F6': 'F5', 'F2': 'F2', 'F4': 'F3', 'F8': 'F1'}

{'C1': 'C1', 'C2': 'C2'}

player B win

{'C1': 'player B lose', 'C2': 'player B win'}

SVMClassifier_poly

C1

Employee Attrition

The class assigned by the model is C1 with a close to 97.67% confidence level, implying that the likelihood of C2 is only 2.33%. Based on the analysis, the most important features considered during the classification are F5, F30, F15, and F10 but among these features, F30 and F15 are the only ones with negative attributions, decreasing the likelihood of C1 being the label for the given case. Furthermore, moderately influencing the decision are F28, F7, F2, and F11. F28, F7, and F2 have positive attributions, while F11 has a negative impact, shifting the prediction in a different direction. Finally, the features with insignificant impact on the model when it comes to this case include F16, F13, F26, and F22.

{'C1': '97.67%', 'C2': '2.33%'}

{'F5': 'OverTime', 'F30': 'JobSatisfaction', 'F15': 'BusinessTravel', 'F10': 'MaritalStatus', 'F28': 'EnvironmentSatisfaction', 'F7': 'Department', 'F2': 'Age', 'F11': 'YearsInCurrentRole', 'F23': 'TotalWorkingYears', 'F12': 'WorkLifeBalance', 'F17': 'JobLevel', 'F18': 'JobInvolvement', 'F8': 'EducationField', 'F1': 'JobRole', 'F9': 'MonthlyIncome', 'F6': 'PerformanceRating', 'F20': 'DistanceFromHome', 'F19': 'Education', 'F25': 'Gender', 'F29': 'YearsWithCurrManager', 'F16': 'PercentSalaryHike', 'F13': 'RelationshipSatisfaction', 'F26': 'MonthlyRate', 'F22': 'DailyRate', 'F4': 'YearsSinceLastPromotion', 'F3': 'HourlyRate', 'F21': 'YearsAtCompany', 'F14': 'TrainingTimesLastYear', 'F24': 'StockOptionLevel', 'F27': 'NumCompaniesWorked'}

{'F26': 'F5', 'F30': 'F30', 'F17': 'F15', 'F25': 'F10', 'F28': 'F28', 'F21': 'F7', 'F1': 'F2', 'F14': 'F11', 'F11': 'F23', 'F20': 'F12', 'F5': 'F17', 'F29': 'F18', 'F22': 'F8', 'F24': 'F1', 'F6': 'F9', 'F19': 'F6', 'F3': 'F20', 'F27': 'F19', 'F23': 'F25', 'F16': 'F29', 'F9': 'F16', 'F18': 'F13', 'F7': 'F26', 'F2': 'F22', 'F15': 'F4', 'F4': 'F3', 'F13': 'F21', 'F12': 'F14', 'F10': 'F24', 'F8': 'F27'}

{'C2': 'C1', 'C1': 'C2'}

Stay

{'C1': 'Leave', 'C2': 'Leave'}

KNeighborsClassifier

C2

Advertisement Prediction

With a higher degree of confidence, the model labels this given case as C2 since there is a zero chance that it is C1. The classification here can be attributed to all the features having positive contributions, decreasing the odds of C1 being the correct label. The features can be ranked based on their degree of influence from the most relevant to the least relevant as follows: F1, F7, F5, F6, F4, F2, F3. This implies that F1 is the most influential feature, while F3 is the least influential among the input features.

{'C1': '0.00%', 'C2': '100.00%'}

{'F1': 'Daily Time Spent on Site', 'F7': 'Area Income', 'F5': 'Age', 'F6': 'Daily Internet Usage', 'F4': 'ad_day', 'F2': 'Gender', 'F3': 'ad_month'}

{'F1': 'F1', 'F3': 'F7', 'F2': 'F5', 'F4': 'F6', 'F7': 'F4', 'F5': 'F2', 'F6': 'F3'}

{'C2': 'C1', 'C1': 'C2'}

Watch

{'C1': 'Skip', 'C2': 'Watch'}

SVM_poly

C1

Mobile Price-Range Classification

According to the model, C1 has a prediction probability of 99.45 percent, C3 has a prediction probability of 0.47 percent, C4 has a prediction probability of 0.04 percent, and C2 has a prediction probability of 0.05 percent, therefore, the most likely class is C1. F19 and F20 positively influence the above-mentioned label decision in favour of C1, but F1 has the opposite effect, favouring a different label. F13 and F8 both have a similar negative impact on the C1 prediction, whereas F2 has a positive impact. In this case, F5, F15, F11, and F16 have little influence on the labelling result. All in all, the model is confident in its assignment of the C1 class as shown by the predicted probabilities across the classes.

{'C1': '99.45%', 'C3': '0.47%', 'C4': '0.04%', 'C2': '0.05%'}

{'F20': 'ram', 'F19': 'battery_power', 'F1': 'px_height', 'F13': 'px_width', 'F2': 'dual_sim', 'F8': 'four_g', 'F5': 'touch_screen', 'F15': 'int_memory', 'F11': 'pc', 'F16': 'n_cores', 'F9': 'fc', 'F18': 'clock_speed', 'F12': 'three_g', 'F4': 'sc_w', 'F6': 'wifi', 'F3': 'm_dep', 'F7': 'mobile_wt', 'F10': 'talk_time', 'F14': 'sc_h', 'F17': 'blue'}

{'F11': 'F20', 'F1': 'F19', 'F9': 'F1', 'F10': 'F13', 'F16': 'F2', 'F17': 'F8', 'F19': 'F5', 'F4': 'F15', 'F8': 'F11', 'F7': 'F16', 'F3': 'F9', 'F2': 'F18', 'F18': 'F12', 'F13': 'F4', 'F20': 'F6', 'F5': 'F3', 'F6': 'F7', 'F14': 'F10', 'F12': 'F14', 'F15': 'F17'}

{'C1': 'C1', 'C2': 'C3', 'C4': 'C4', 'C3': 'C2'}

r1

{'C1': 'r1', 'C3': 'r2', 'C4': 'r3', 'C2': 'r4'}

GradientBoostingClassifier

C2

Food Ordering Customer Churn Prediction

Per the model employed here, the prediction probability of C1 is only 17.93%, and that of C2 is equal to 82.07%. Given the information provided to the model, the most valid conclusion regarding the true label is that C2 is without a doubt the most likely one. The attributions analysis indicates that F9, F16, F22, F4, and F23 are the major drivers resulting in the prediction probabilities across the classes under consideration. At the tail end are features such as F42, F25, F24, and F15 that have very little influence on the decision made with respect to the given case. Among the influential features, only F9, F16, F4, F10, F34, F35, F29, F5, and F19 have positive contributions in support of labelling the given case as C2. On the other hand, the negative features such as F22, F23, F45, F3, F36, F20, and F26, suggest C1 could likely be the true label in this case. Overall, the marginal doubt in the correctness of assigning C2 to the case under consideration is attributed to the negative features driving the model's decision in the direction of C1 away from C2. But the higher influence of positive features such as F9 and F16 ensures that C2 is assigned as the most probable label.

{'C2': '82.07%', 'C1': '17.93%'}

{'F9': 'Ease and convenient', 'F16': 'More restaurant choices', 'F22': 'Bad past experience', 'F4': 'More Offers and Discount', 'F23': 'Unavailability', 'F45': 'Good Food quality', 'F10': 'Low quantity low time', 'F34': 'Delay of delivery person getting assigned', 'F3': 'Late Delivery', 'F35': 'Less Delivery time', 'F26': 'Residence in busy location', 'F36': 'Freshness ', 'F20': 'Educational Qualifications', 'F29': 'Influence of rating', 'F32': 'Occupation', 'F44': 'Perference(P1)', 'F43': 'Delivery person ability', 'F5': 'Good Taste ', 'F19': 'Long delivery time', 'F14': 'Self Cooking', 'F42': 'Influence of time', 'F25': 'High Quality of package', 'F24': 'Number of calls', 'F15': 'Good Road Condition', 'F37': 'Politeness', 'F2': 'Google Maps Accuracy', 'F8': 'Temperature', 'F1': 'Maximum wait time', 'F33': 'Order Time', 'F46': 'Age', 'F17': 'Order placed by mistake', 'F27': 'Missing item', 'F39': 'Wrong order delivered', 'F38': 'Delay of delivery person picking up food', 'F7': 'Family size', 'F13': 'Unaffordable', 'F40': 'Poor Hygiene', 'F18': 'Health Concern', 'F41': 'Good Tracking system', 'F21': 'Easy Payment option', 'F31': 'Time saving', 'F11': 'Perference(P2)', 'F12': 'Monthly Income', 'F6': 'Marital Status', 'F30': 'Gender', 'F28': 'Good Quantity'}

{'F10': 'F9', 'F12': 'F16', 'F21': 'F22', 'F14': 'F4', 'F22': 'F23', 'F15': 'F45', 'F36': 'F10', 'F25': 'F34', 'F19': 'F3', 'F39': 'F35', 'F33': 'F26', 'F43': 'F36', 'F6': 'F20', 'F38': 'F29', 'F4': 'F32', 'F8': 'F44', 'F37': 'F43', 'F45': 'F5', 'F24': 'F19', 'F17': 'F14', 'F30': 'F42', 'F40': 'F25', 'F41': 'F24', 'F35': 'F15', 'F42': 'F37', 'F34': 'F2', 'F44': 'F8', 'F32': 'F1', 'F31': 'F33', 'F1': 'F46', 'F29': 'F17', 'F28': 'F27', 'F27': 'F39', 'F26': 'F38', 'F7': 'F7', 'F23': 'F13', 'F20': 'F40', 'F18': 'F18', 'F16': 'F41', 'F13': 'F21', 'F11': 'F31', 'F9': 'F11', 'F5': 'F12', 'F3': 'F6', 'F2': 'F30', 'F46': 'F28'}

{'C1': 'C2', 'C2': 'C1'}

Return

{'C2': 'Return', 'C1': 'Go Away'}

RandomForestClassifier

C1

Personal Loan Modelling

The model is about 90.0% certain or sure that the correct label based on the input features of the given case is C1. The features with the most significant influence on the decision are F9, F5, F7, and F4. The influence of the features can be categorised as positive or negative traits depending on the direction of the effect on the model. Positive features increase the likelihood of the most likely class (i.e., C1), whereas negative features reduce the model's responsiveness to the assigned label, favouring the less likely class (i.e., C2). From the attribution analysis, F3, F2, and F1 are the negative features here. Overall, the negative features are shown to have moderate to low influence compared to the positive features, hence explaining why the model is very confident about the assigned label C1.

{'C2': '10.00%', 'C1': '90.00%'}

{'F5': 'Income', 'F9': 'CCAvg', 'F7': 'CD Account', 'F4': 'Education', 'F2': 'Extra_service', 'F8': 'Securities Account', 'F6': 'Family', 'F3': 'Mortgage', 'F1': 'Age'}

{'F2': 'F5', 'F4': 'F9', 'F8': 'F7', 'F5': 'F4', 'F9': 'F2', 'F7': 'F8', 'F3': 'F6', 'F6': 'F3', 'F1': 'F1'}

{'C2': 'C2', 'C1': 'C1'}

Accept

{'C2': 'Reject', 'C1': 'Accept'}

LogisticRegression

C1

Tic-Tac-Toe Strategy

With an 81.01% chance of being correct, C1 is the most likely label, consequently, the C2 class's prediction probability is only 18.99%. The algorithm or classifier got the above prediction mostly due to the influence of features like F6, F2, F4, and F7. F5, which is found to have very little impact with regard to the label choice here, is the least relevant feature for the algorithm. F2, F1, F4, and F7 have a positive direction of influence, pushing the algorithm higher towards the C1 label. Negative features like F6, F3, and F8 favour choosing or labelling the case as C2.

{'C2': '18.99%', 'C1': '81.01%'}

{'F2': 'bottom-right-square', 'F6': 'middle-middle-square', 'F4': 'bottom-left-square', 'F7': 'middle-left-square', 'F1': 'top-left-square', 'F8': ' top-right-square', 'F3': 'middle-right-square', 'F9': 'top-middle-square', 'F5': 'bottom-middle-square'}

{'F9': 'F2', 'F5': 'F6', 'F7': 'F4', 'F4': 'F7', 'F1': 'F1', 'F3': 'F8', 'F6': 'F3', 'F2': 'F9', 'F8': 'F5'}

{'C2': 'C2', 'C1': 'C1'}

player B win

{'C2': 'player B lose', 'C1': 'player B win'}

SVC

C1

Student Job Placement

The model makes classification decisions based on the information provided to it and for the case here, the prediction probabilities across the two class labels, C2 and C1, are 49.32% and 50.68%, respectively. Based on these prediction probabilities, the label assigned is C1, since it has the highest likelihood, however, the model is not very certain about the correctness of the assigned label since its probability is marginally higher than the average. The uncertainty in the classification here can be blamed on the fact that only F3, F10, F7, F9, and F2 have positive attributions, shifting the decision higher towards C1. On the other hand, features F12, F5, F6, F11, F8, F4, and F1 have negative contributions that decrease the prediction likelihood of C1 while increasing that of C2. To cut a long story short, the most positive features are F3 and F10, whereas the most negative ones are F12 and F5. Finally, F8, F2, and F4 are not as important as all the previously mentioned features hence received little attention from the model.

{'C2': '49.32%', 'C1': '50.68%'}

{'F3': 'mba_p', 'F12': 'specialisation', 'F5': 'etest_p', 'F10': 'gender', 'F6': 'workex', 'F7': 'hsc_s', 'F11': 'hsc_p', 'F9': 'degree_t', 'F1': 'ssc_p', 'F8': 'degree_p', 'F2': 'ssc_b', 'F4': 'hsc_b'}

{'F5': 'F3', 'F12': 'F12', 'F4': 'F5', 'F6': 'F10', 'F11': 'F6', 'F9': 'F7', 'F2': 'F11', 'F10': 'F9', 'F1': 'F1', 'F3': 'F8', 'F7': 'F2', 'F8': 'F4'}

{'C2': 'C2', 'C1': 'C1'}

Placed

{'C2': 'Not Placed', 'C1': 'Placed'}

SVMClassifier_liner

C2

Employee Attrition

The most likely label for the given case is C2 since the predicted probability of C1 is only 34.27% and this means that the likelihood of C2 is 65.73%. The most relevant features that led to the C2 classification verdict are F5, F29, F30, F14, and F6. However, some of the features are deemed irrelevant to the above verdict and these include F28, F4, F9, and F18. Among the relevant features with some degree of impact, seven are shown to drive the model's class assignment towards the C1, while the remaining support the C2 prediction. Notable negative features swinging the prediction towards C1 are F5, F29, and F30, while the notable positive features are F14 and F6. The small uncertainty associated with the prediction decision for the given case could be attributed to the fact that all the three most important features are negative features whose values contradict assigning the label C2.

{'C2': '65.73%', 'C1': '34.27%'}

{'F5': 'OverTime', 'F29': 'NumCompaniesWorked', 'F30': 'YearsSinceLastPromotion', 'F14': 'BusinessTravel', 'F6': 'MaritalStatus', 'F2': 'RelationshipSatisfaction', 'F15': 'Department', 'F13': 'Age', 'F3': 'Gender', 'F20': 'JobInvolvement', 'F19': 'JobRole', 'F1': 'PerformanceRating', 'F8': 'EnvironmentSatisfaction', 'F10': 'DailyRate', 'F16': 'YearsAtCompany', 'F24': 'YearsWithCurrManager', 'F26': 'Education', 'F23': 'EducationField', 'F21': 'WorkLifeBalance', 'F7': 'DistanceFromHome', 'F28': 'YearsInCurrentRole', 'F4': 'TrainingTimesLastYear', 'F9': 'TotalWorkingYears', 'F18': 'StockOptionLevel', 'F27': 'PercentSalaryHike', 'F12': 'MonthlyRate', 'F22': 'MonthlyIncome', 'F11': 'JobLevel', 'F17': 'HourlyRate', 'F25': 'JobSatisfaction'}

{'F26': 'F5', 'F8': 'F29', 'F15': 'F30', 'F17': 'F14', 'F25': 'F6', 'F18': 'F2', 'F21': 'F15', 'F1': 'F13', 'F23': 'F3', 'F29': 'F20', 'F24': 'F19', 'F19': 'F1', 'F28': 'F8', 'F2': 'F10', 'F13': 'F16', 'F16': 'F24', 'F27': 'F26', 'F22': 'F23', 'F20': 'F21', 'F3': 'F7', 'F14': 'F28', 'F12': 'F4', 'F11': 'F9', 'F10': 'F18', 'F9': 'F27', 'F7': 'F12', 'F6': 'F22', 'F5': 'F11', 'F4': 'F17', 'F30': 'F25'}

{'C1': 'C2', 'C2': 'C1'}

Stay

{'C2': 'Leave', 'C1': 'Leave'}

RandomForestClassifier

C2

Printer Sales

Per the classifier for the given data, the most plausible label is C2. F5, F10, F13, and F4 are the main features pushing for the above-mentioned outcome. F24, F2, F25, F12, F16, and F1, on the other hand, have little contribution to the classifier employed here. F8, F19, F17, and F22 have a moderate contribution to the assignment of C2. The classifier's confidence in the label decision above can be attributed to larger positive attributions of F19, F8, F13, and F10 compared to the negative attributions of F17, F3, F5, F23, F4, and F15.

{'C1': '20.00%', 'C2': '80.00%'}

{'F10': 'X24', 'F13': 'X1', 'F5': 'X8', 'F4': 'X21', 'F17': 'X4', 'F19': 'X10', 'F8': 'X3', 'F22': 'X15', 'F14': 'X9', 'F26': 'X23', 'F11': 'X25', 'F9': 'X7', 'F23': 'X22', 'F7': 'X11', 'F3': 'X17', 'F20': 'X18', 'F21': 'X26', 'F15': 'X13', 'F18': 'X6', 'F6': 'X20', 'F12': 'X16', 'F1': 'X19', 'F25': 'X2', 'F24': 'X12', 'F16': 'X5', 'F2': 'X14'}

{'F24': 'F10', 'F1': 'F13', 'F8': 'F5', 'F21': 'F4', 'F4': 'F17', 'F10': 'F19', 'F3': 'F8', 'F15': 'F22', 'F9': 'F14', 'F23': 'F26', 'F25': 'F11', 'F7': 'F9', 'F22': 'F23', 'F11': 'F7', 'F17': 'F3', 'F18': 'F20', 'F26': 'F21', 'F13': 'F15', 'F6': 'F18', 'F20': 'F6', 'F16': 'F12', 'F19': 'F1', 'F2': 'F25', 'F12': 'F24', 'F5': 'F16', 'F14': 'F2'}

{'C1': 'C1', 'C2': 'C2'}

More

{'C1': 'Less', 'C2': 'More'}

SGDClassifier

C2

Job Change of Data Scientists

The least probable class, according to the classification algorithm, is C1, with a prediction probability of 25.12%, therefore, we can conclude that the algorithm is quite confident that the correct label for this data is C2. Analysing the attributions revealed that F3, F4, F8, and F6 are the most relevant features, whereas F11, F5, and F1 are the least relevant features. Increasing the algorithm's response in favour of C2 are the positive features F3, F8, F6, F5, F11, and F12. On the contrary, all the other features, F4, F9, F7, F2, F10, and F1, drive the algorithm towards labelling the given data as C1, hence they are considered negative features. Furthermore, the negative influence on the algorithm is the reason why the confidence level in the C2 is reduced to 74.88%.

{'C2': '74.88%', 'C1': '25.12%'}

{'F3': 'city_development_index', 'F8': 'relevent_experience', 'F4': 'city', 'F6': 'major_discipline', 'F9': 'experience', 'F7': 'training_hours', 'F2': 'education_level', 'F12': 'gender', 'F10': 'enrolled_university', 'F11': 'company_type', 'F5': 'last_new_job', 'F1': 'company_size'}

{'F1': 'F3', 'F5': 'F8', 'F3': 'F4', 'F8': 'F6', 'F9': 'F9', 'F2': 'F7', 'F7': 'F2', 'F4': 'F12', 'F6': 'F10', 'F11': 'F11', 'F12': 'F5', 'F10': 'F1'}

{'C1': 'C2', 'C2': 'C1'}

Stay

{'C2': 'Stay', 'C1': 'Leave'}

SVM_poly

C2

Mobile Price-Range Classification

According to the classification algorithm, neither C1 nor C3 nor C4 is the correct label for the given case. It is 100.0% certain that C2 is the right label. The higher degree of certainty in the above prediction can be attributed to the positive contributions of F5, F16, and F12. The other positive features include F13, F18, F4, and F11, however, unlike F5, F16, and F12, these features have a moderately low impact on the algorithm's decision. The remaining positive features, F6, F15, F1, and F14, are among the least influential input features considered by the algorithm. There are other features such as F10, F9, F17, and F20 whose contributions only serve to decrease the odds of C2 being the correct label for the given case. Regarding the high confidence of the algorithm with respect to this classification, one can conclude that the negative features have little influence on the algorithm's label decision here.

{'C1': '0.00%', 'C3': '0.00%', 'C4': '0.00%', 'C2': '100.00%'}

{'F5': 'ram', 'F16': 'battery_power', 'F12': 'px_width', 'F10': 'int_memory', 'F9': 'sc_h', 'F17': 'wifi', 'F13': 'fc', 'F20': 'three_g', 'F18': 'mobile_wt', 'F19': 'clock_speed', 'F3': 'm_dep', 'F8': 'n_cores', 'F4': 'pc', 'F2': 'touch_screen', 'F11': 'blue', 'F6': 'talk_time', 'F7': 'sc_w', 'F15': 'px_height', 'F1': 'four_g', 'F14': 'dual_sim'}

{'F11': 'F5', 'F1': 'F16', 'F10': 'F12', 'F4': 'F10', 'F12': 'F9', 'F20': 'F17', 'F3': 'F13', 'F18': 'F20', 'F6': 'F18', 'F2': 'F19', 'F5': 'F3', 'F7': 'F8', 'F8': 'F4', 'F19': 'F2', 'F15': 'F11', 'F14': 'F6', 'F13': 'F7', 'F9': 'F15', 'F17': 'F1', 'F16': 'F14'}

{'C1': 'C1', 'C3': 'C3', 'C2': 'C4', 'C4': 'C2'}

r4

{'C1': 'r1', 'C3': 'r2', 'C4': 'r3', 'C2': 'r4'}

DNN

C1

Ethereum Fraud Detection

The prediction likelihoods across the two classes are 15.35% for class C2 and 84.65% for C1, it can be concluded that C1 is the most probable class label for the given data instance. According to the attribution analysis conducted, the different input variables have varying degrees of influence on the model's decision here. The most influential set of variables is F11, F35, F31, F5, F1, F9, and F33, while the variables with the least influence include F8, F10, F34, F19, F4, and F6. The following or subsequent analysis performed to understand the direction of contribution of of the features will focus on the most influential ones controlling the label selection here. Among the top influential features, F11, F35, F31, F5, and F33, only F11 and F35 have negative contributions, decreasing the probability that C1 is the correct label, and they strongly support labelling the case as C2 instead. Pushing the classification decision in favour of C1 are the positive variables such as F31, F5, and F33. The contributions of the remaining variables, including F1, F9, and F14, have moderate to low influence. All in all, the marginal uncertainty in the decision here is mainly due to the negative influences of F11, F35, F18, and F32, but the positive contributions of F31, F5, F14, F1, F9, and F33 drive the decision higher towards C1.

{'C2': '15.35%', 'C1': '84.65%'}

{'F11': ' ERC20 uniq rec contract addr', 'F35': ' ERC20 uniq rec token name', 'F31': 'min value received', 'F5': 'Time Diff between first and last (Mins)', 'F33': 'avg val sent', 'F1': ' ERC20 uniq sent token name', 'F9': 'Sent tnx', 'F14': 'Avg min between received tnx', 'F18': 'Unique Received From Addresses', 'F32': ' ERC20 uniq rec addr', 'F26': 'total transactions (including tnx to create contract', 'F20': 'Avg min between sent tnx', 'F15': ' ERC20 uniq sent addr.1', 'F7': 'avg val received', 'F27': 'Unique Sent To Addresses', 'F23': 'max value received ', 'F28': 'max val sent', 'F37': 'min val sent', 'F22': 'Number of Created Contracts', 'F38': 'total ether received', 'F16': ' ERC20 uniq sent addr', 'F12': ' ERC20 total Ether received', 'F17': 'Received Tnx', 'F29': ' ERC20 avg val sent', 'F2': 'total Ether sent', 'F13': ' ERC20 min val sent', 'F21': 'max val sent to contract', 'F24': 'total ether balance', 'F30': ' ERC20 max val sent', 'F3': ' Total ERC20 tnxs', 'F25': ' ERC20 total ether sent', 'F36': ' ERC20 avg val rec', 'F8': 'avg value sent to contract', 'F10': ' ERC20 min val rec', 'F34': ' ERC20 max val rec', 'F19': ' ERC20 total Ether sent contract', 'F4': 'min value sent to contract', 'F6': 'total ether sent contracts'}

{'F30': 'F11', 'F38': 'F35', 'F9': 'F31', 'F3': 'F5', 'F14': 'F33', 'F37': 'F1', 'F4': 'F9', 'F2': 'F14', 'F7': 'F18', 'F28': 'F32', 'F18': 'F26', 'F1': 'F20', 'F29': 'F15', 'F11': 'F7', 'F8': 'F27', 'F10': 'F23', 'F13': 'F28', 'F12': 'F37', 'F6': 'F22', 'F20': 'F38', 'F27': 'F16', 'F24': 'F12', 'F5': 'F17', 'F36': 'F29', 'F19': 'F2', 'F34': 'F13', 'F16': 'F21', 'F22': 'F24', 'F35': 'F30', 'F23': 'F3', 'F25': 'F25', 'F33': 'F36', 'F17': 'F8', 'F31': 'F10', 'F32': 'F34', 'F26': 'F19', 'F15': 'F4', 'F21': 'F6'}

{'C1': 'C2', 'C2': 'C1'}

Fraud

{'C2': 'Not Fraud', 'C1': 'Fraud'}

KNeighborsClassifier

C1

Credit Risk Classification

According to the machine learning model, it is more likely that the case's label is C1, with a certainty of 100.0%, and this prediction decision is mainly based on the effects of the following features: F8, F9, F10, F4, and F1 on the model. Apart from F1 and F4, all the other variables mentioned above have a strong positive influence, improving the odds of the prediction class, C1. Together with F1 and F4, the values of variables F11 and F6 indicate that C2 could be the correct label instead. Unlike the top positive variables, F8, F9, and F10, each of these negative variables has a moderate contribution to the final decision. The features F5, F7, F3, and F2 are shown to have made minor contributions to the model's decision in this case. In summary, with only the positive contributions from F8, F9, F10, F3, and F5, the model is very certain of the classification output as indicated by the predicted probabilities across C1 and C2.

{'C1': '100.00%', 'C2': '0.00%'}

{'F8': 'fea_4', 'F9': 'fea_8', 'F10': 'fea_2', 'F1': 'fea_9', 'F4': 'fea_6', 'F11': 'fea_10', 'F6': 'fea_1', 'F5': 'fea_7', 'F7': 'fea_11', 'F2': 'fea_3', 'F3': 'fea_5'}

{'F4': 'F8', 'F8': 'F9', 'F2': 'F10', 'F9': 'F1', 'F6': 'F4', 'F10': 'F11', 'F1': 'F6', 'F7': 'F5', 'F11': 'F7', 'F3': 'F2', 'F5': 'F3'}

{'C2': 'C1', 'C1': 'C2'}

Low

{'C1': 'Low', 'C2': 'High'}

SVMClassifier_poly

C2

Employee Attrition

The classification findings by the model for the case here are as follows: there is a 97.67% chance that C2 is the correct label hence only a marginally low chance of 2.33% that C2 is not the correct label but C1 is. From the above findings, it is valid to conclude that the right class for the given case is C2, and the model is very certain of this decision. The features with the most control and influence on the classification above are F13, F28, F2, F22, and F14 but the influence of the remaining features is either moderate or low or negligible. Some of the features with moderate impact include F6, F24, F3, and F27. Those with low influence are F26, F17, F23, F16, and F29. Finally, those with negligible impact are F5, F21, F30, F1, F7, F12, F8, F15, F18, and F19 since their values are shown to have no impact on the classification made by the model here. The top positive features increasing the prediction likelihood of class C2 are F13, F14, and F20. Conversely, the negative features decreasing the odds in favour of C1 are primarily F28, F3, and F2.

{'C2': '97.67%', 'C1': '2.33%'}

{'F13': 'OverTime', 'F28': 'JobSatisfaction', 'F2': 'BusinessTravel', 'F22': 'MaritalStatus', 'F14': 'EnvironmentSatisfaction', 'F6': 'Department', 'F24': 'Age', 'F3': 'YearsInCurrentRole', 'F27': 'TotalWorkingYears', 'F11': 'WorkLifeBalance', 'F20': 'JobLevel', 'F9': 'JobInvolvement', 'F10': 'EducationField', 'F25': 'JobRole', 'F4': 'MonthlyIncome', 'F26': 'PerformanceRating', 'F17': 'DistanceFromHome', 'F23': 'Education', 'F16': 'Gender', 'F29': 'YearsWithCurrManager', 'F5': 'PercentSalaryHike', 'F21': 'RelationshipSatisfaction', 'F30': 'MonthlyRate', 'F1': 'DailyRate', 'F7': 'YearsSinceLastPromotion', 'F12': 'HourlyRate', 'F8': 'YearsAtCompany', 'F15': 'TrainingTimesLastYear', 'F18': 'StockOptionLevel', 'F19': 'NumCompaniesWorked'}

{'F26': 'F13', 'F30': 'F28', 'F17': 'F2', 'F25': 'F22', 'F28': 'F14', 'F21': 'F6', 'F1': 'F24', 'F14': 'F3', 'F11': 'F27', 'F20': 'F11', 'F5': 'F20', 'F29': 'F9', 'F22': 'F10', 'F24': 'F25', 'F6': 'F4', 'F19': 'F26', 'F3': 'F17', 'F27': 'F23', 'F23': 'F16', 'F16': 'F29', 'F9': 'F5', 'F18': 'F21', 'F7': 'F30', 'F2': 'F1', 'F15': 'F7', 'F4': 'F12', 'F13': 'F8', 'F12': 'F15', 'F10': 'F18', 'F8': 'F19'}

{'C1': 'C2', 'C2': 'C1'}

Stay

{'C2': 'Leave', 'C1': 'Leave'}

LogisticRegression

C2

Flight Price-Range Classification

The model is very confident that C2 is the most probable class for the given case, with a probability of 90.48% which means that the other labels are very unlikely. F1 and F8 are the most important variables with respect to this classification verdict while all other variables are shown to have a medium or low impact. Fortunately, the top variables, F1 and F8, have the same direction of influence, increasing the likelihood of C2. Furthermore, while F5 and F10 push the model to predict C2, those pushing for the assignment of a different label are F2, F7, and F3. Finally, many features have a fairly small impact on the final prediction made by the model here, but F12, F7, and F11 have the least impact.

{'C2': '90.48%', 'C3': '9.51%', 'C1': '0.01%'}

{'F1': 'Total_Stops', 'F8': 'Airline', 'F10': 'Destination', 'F5': 'Arrival_hour', 'F2': 'Source', 'F4': 'Duration_hours', 'F6': 'Dep_hour', 'F3': 'Dep_minute', 'F9': 'Arrival_minute', 'F12': 'Journey_month', 'F7': 'Journey_day', 'F11': 'Duration_mins'}

{'F12': 'F1', 'F9': 'F8', 'F11': 'F10', 'F5': 'F5', 'F10': 'F2', 'F7': 'F4', 'F3': 'F6', 'F4': 'F3', 'F6': 'F9', 'F2': 'F12', 'F1': 'F7', 'F8': 'F11'}

{'C3': 'C2', 'C1': 'C3', 'C2': 'C1'}

Low

{'C2': 'Low', 'C3': 'Moderate', 'C1': 'High'}

SVC

C2

Water Quality Classification

Despite the reasonably high confidence in the assigned label, the prediction probabilities across the two classes indicate that C1 might be the correct label. F7, F6, F9, and F3 are the factors whose major contributions resulted in the labelling choice mentioned above. According to the analysis, the top two factors, F7 and F6, have a negative influence, leading the classifier to classify the data as C1 rather than C2. F8 is the only other negative variable with a moderate effect when compared to the other two negative variables. Nevertheless, there are several factors, F9, F3, F2, F4, F1, and F5, that favourably support and encourage the classifier to assign C2. All in all, the degree of uncertainty in this classification instance might be explained by just looking at the negative factors' rather strong pull on the classifier towards C1.

{'C1': '38.68%', 'C2': '61.32%'}

{'F7': 'Sulfate', 'F6': 'Hardness', 'F9': 'ph', 'F3': 'Conductivity', 'F8': 'Turbidity', 'F2': 'Chloramines', 'F4': 'Solids', 'F1': 'Trihalomethanes', 'F5': 'Organic_carbon'}

{'F5': 'F7', 'F2': 'F6', 'F1': 'F9', 'F6': 'F3', 'F9': 'F8', 'F4': 'F2', 'F3': 'F4', 'F8': 'F1', 'F7': 'F5'}

{'C2': 'C1', 'C1': 'C2'}

Portable

{'C1': 'Not Portable', 'C2': 'Portable'}

MLPClassifier

C2

Ethereum Fraud Detection

C1 has a probability estimate of only 6.80%, while that of C2 is 93.20%; consequently, the most likely class for the given case is C2. The important or relevant features considered by the classifier are F29, F17, F9, F33, F26, F32, F8, F1, F20, F19, F3, F34, F14, F7, F4, F36, F25, F2, F12, and F23. Not all input features are relevant when determining the appropriate label and these irrelevant features include F5, F21, and F11. Furthermore, F29 and F17 have a strong positive effect, increasing the odds in favour of C2. In contrast, the F9, F26, and F33 are the negative features, lowering the odds of C2. Comparing the attributions of F29, F32, and F17 features to those of the negative features mentioned above, it is not surprising that the classifier is convinced that C2 is the most likely label here.

{'C1': '6.80%', 'C2': '93.20%'}

{'F29': 'Unique Received From Addresses', 'F17': ' ERC20 total Ether sent contract', 'F9': 'total ether received', 'F33': 'Sent tnx', 'F26': 'Number of Created Contracts', 'F32': ' ERC20 uniq rec token name', 'F8': ' ERC20 uniq rec contract addr', 'F1': 'max value received ', 'F20': 'total transactions (including tnx to create contract', 'F19': ' ERC20 uniq sent addr.1', 'F3': ' ERC20 uniq sent addr', 'F34': 'Received Tnx', 'F14': 'avg val received', 'F7': ' ERC20 uniq rec addr', 'F4': 'avg val sent', 'F36': 'min value received', 'F25': 'Unique Sent To Addresses', 'F2': ' ERC20 uniq sent token name', 'F12': 'Avg min between received tnx', 'F23': 'Time Diff between first and last (Mins)', 'F5': ' ERC20 min val rec', 'F11': ' ERC20 max val rec', 'F21': ' ERC20 min val sent', 'F13': ' ERC20 max val sent', 'F28': ' ERC20 avg val sent', 'F35': ' ERC20 avg val rec', 'F30': ' Total ERC20 tnxs', 'F38': ' ERC20 total ether sent', 'F27': ' ERC20 total Ether received', 'F6': 'total ether balance', 'F24': 'total ether sent contracts', 'F10': 'total Ether sent', 'F18': 'avg value sent to contract', 'F22': 'max val sent to contract', 'F15': 'min value sent to contract', 'F37': 'max val sent', 'F31': 'min val sent', 'F16': 'Avg min between sent tnx'}

{'F7': 'F29', 'F26': 'F17', 'F20': 'F9', 'F4': 'F33', 'F6': 'F26', 'F38': 'F32', 'F30': 'F8', 'F10': 'F1', 'F18': 'F20', 'F29': 'F19', 'F27': 'F3', 'F5': 'F34', 'F11': 'F14', 'F28': 'F7', 'F14': 'F4', 'F9': 'F36', 'F8': 'F25', 'F37': 'F2', 'F2': 'F12', 'F3': 'F23', 'F31': 'F5', 'F32': 'F11', 'F34': 'F21', 'F35': 'F13', 'F36': 'F28', 'F33': 'F35', 'F23': 'F30', 'F25': 'F38', 'F24': 'F27', 'F22': 'F6', 'F21': 'F24', 'F19': 'F10', 'F17': 'F18', 'F16': 'F22', 'F15': 'F15', 'F13': 'F37', 'F12': 'F31', 'F1': 'F16'}

{'C1': 'C1', 'C2': 'C2'}

Fraud

{'C1': 'Not Fraud', 'C2': 'Fraud'}

BernoulliNB

C2

German Credit Evaluation

The model is not 100% convinced that the correct label for the data under consideration is C2 since there is a 26.27% chance that labelling the data as C1 is correct. All the input variables are shown to have some degree of influence on the classification decision, with the most influential variables being F2, F8, and F3, whereas F1 and F9 are the least influential. The impact of F6, F5, F7, and F4 can be considered moderate compared to the F2, F8, and F3. The uncertainty surrounding the above classification can be blamed on the fact that the majority of input variables have values suggesting that C1 could be the appropriate label. The negative features that decrease the prediction likelihood of C2 are F2, F3, F7, and F4. However, given that the prediction probability is about 73.73%, it can be said that the influence of positive features, F8, F6, F5, and F1, is enough to swing the model's verdict in favour of C2.

{'C2': '73.73%', 'C1': '26.27%'}

{'F2': 'Saving accounts', 'F8': 'Sex', 'F3': 'Housing', 'F6': 'Purpose', 'F5': 'Checking account', 'F7': 'Job', 'F4': 'Duration', 'F1': 'Age', 'F9': 'Credit amount'}

{'F5': 'F2', 'F2': 'F8', 'F4': 'F3', 'F9': 'F6', 'F6': 'F5', 'F3': 'F7', 'F8': 'F4', 'F1': 'F1', 'F7': 'F9'}

{'C2': 'C2', 'C1': 'C1'}

Good Credit

{'C2': 'Good Credit', 'C1': 'Bad Credit'}

SVMClassifier_poly

C1

Employee Attrition

The model predicted class C1 with an 81.98% prediction likelihood. F29 had the largest impact, followed by F18, F19, F17, F3, F1, F21, F8, F4, F20, F24, F23, F25, F10, F28, F27, F11, F30, F9, and finally, F5, which had the smallest non-zero impact. F29, the feature with the largest impact, contributed against the direction of the prediction, whereas F18, F19, F17, and F3 all contributed positively towards the prediction. Other features that had a negative influence on the prediction included F21 and F8, whereas F1 had a positive influence on the prediction. F13, F16, F7, and F6 are shown to have close to zero attribution in the model's prediction verdict in the given case.

{'C1': '81.98%', 'C2': '18.02%'}

{'F29': 'OverTime', 'F18': 'JobSatisfaction', 'F19': 'MaritalStatus', 'F17': 'Department', 'F3': 'NumCompaniesWorked', 'F1': 'BusinessTravel', 'F21': 'JobRole', 'F8': 'EnvironmentSatisfaction', 'F4': 'YearsInCurrentRole', 'F20': 'JobInvolvement', 'F24': 'WorkLifeBalance', 'F23': 'YearsSinceLastPromotion', 'F25': 'TotalWorkingYears', 'F10': 'JobLevel', 'F28': 'Age', 'F27': 'EducationField', 'F11': 'PerformanceRating', 'F30': 'MonthlyRate', 'F9': 'Education', 'F5': 'MonthlyIncome', 'F13': 'DailyRate', 'F16': 'YearsAtCompany', 'F7': 'RelationshipSatisfaction', 'F6': 'TrainingTimesLastYear', 'F22': 'StockOptionLevel', 'F2': 'Gender', 'F12': 'PercentSalaryHike', 'F15': 'HourlyRate', 'F14': 'DistanceFromHome', 'F26': 'YearsWithCurrManager'}

{'F26': 'F29', 'F30': 'F18', 'F25': 'F19', 'F21': 'F17', 'F8': 'F3', 'F17': 'F1', 'F24': 'F21', 'F28': 'F8', 'F14': 'F4', 'F29': 'F20', 'F20': 'F24', 'F15': 'F23', 'F11': 'F25', 'F5': 'F10', 'F1': 'F28', 'F22': 'F27', 'F19': 'F11', 'F7': 'F30', 'F27': 'F9', 'F6': 'F5', 'F2': 'F13', 'F13': 'F16', 'F18': 'F7', 'F12': 'F6', 'F10': 'F22', 'F23': 'F2', 'F9': 'F12', 'F4': 'F15', 'F3': 'F14', 'F16': 'F26'}

{'C1': 'C1', 'C2': 'C2'}

Stay

{'C1': 'Leave', 'C2': 'Leave'}

SVC

C1

German Credit Evaluation

This case's label has a 70.83 percent chance of being C1 and per the predicted likelihoods across the alternative labels, C3 has a 29.71 percent chance of being the correct label, however, the model is certain that C2 is not the true label. The most important variables are F6, F5, F4, and F9, whereas the remaining influential variables are listed in order of the magnitude of their contributions: F7, F1, F3, F8, and F2. Three of the nine variables have values that push towards the prediction of label C3 while the other attributes are referred to as positive since their values inspire the prediction of class C1. F6, F5, and F4 are the three attributes that have a negative influence on the prediction judgement, pushing it away from C1 towards the label C3. Finally, it is essential to highlight that the cumulative effect of positive attributes is greater than that of negative attributes, F4, F5, and F6.

{'C1': '70.83%', 'C3': '29.17%', 'C2': '0.0%'}

{'F9': 'Checking account', 'F6': 'Duration', 'F5': 'Housing', 'F4': 'Saving accounts', 'F7': 'Sex', 'F1': 'Age', 'F3': 'Purpose', 'F8': 'Job', 'F2': 'Credit amount'}

{'F6': 'F9', 'F8': 'F6', 'F4': 'F5', 'F5': 'F4', 'F2': 'F7', 'F1': 'F1', 'F9': 'F3', 'F3': 'F8', 'F7': 'F2'}

{'C3': 'C1', 'C1': 'C3', 'C2': 'C2'}

Good Credit

{'C1': 'Good Credit', 'C3': 'Bad Credit', 'C2': 'Other'}

SVC

C2

Vehicle Insurance Claims

First of all, the classification decision is solely based on the information or data supplied to the prediction model. According to the model, there is a 61.61% chance that C2 is the true label, and a 38.39% chance that C1 is the true label. Since the predicted probability of C2 is higher than that of C1, it is valid to conclude that C2 is most likely the true label. The main feature responsible for this classification is F30, with a very strong positive influence, driving the model's decision higher towards C2. The next set of relevant features are F16, F29, F11, F14, F31, F32, F17, and F26. Among all the features mentioned above, F16, F11, F14, F32, and F17 have negative contributions that are responsible for the decrease in the probability that C2 is the true label. This implies that the contributions of F29, F31, and F26 combined with that of F30 explain why the model is moderately certain that C2 is the true label.

{'C2': '61.61%', 'C1': '38.39%'}

{'F30': 'incident_severity', 'F16': 'insured_hobbies', 'F29': 'authorities_contacted', 'F11': 'insured_education_level', 'F14': 'umbrella_limit', 'F31': 'insured_relationship', 'F32': 'auto_make', 'F17': 'insured_occupation', 'F26': 'capital-gains', 'F1': 'policy_deductable', 'F9': 'policy_state', 'F3': 'auto_year', 'F15': 'insured_sex', 'F24': 'vehicle_claim', 'F20': 'incident_city', 'F22': 'number_of_vehicles_involved', 'F27': 'insured_zip', 'F23': 'injury_claim', 'F6': 'property_claim', 'F13': 'incident_type', 'F33': 'total_claim_amount', 'F10': 'police_report_available', 'F4': 'property_damage', 'F19': 'incident_state', 'F2': 'policy_annual_premium', 'F12': 'incident_hour_of_the_day', 'F25': 'collision_type', 'F7': 'capital-loss', 'F28': 'bodily_injuries', 'F21': 'policy_csl', 'F8': 'witnesses', 'F18': 'age', 'F5': 'months_as_customer'}

{'F27': 'F30', 'F23': 'F16', 'F28': 'F29', 'F21': 'F11', 'F5': 'F14', 'F24': 'F31', 'F33': 'F32', 'F22': 'F17', 'F7': 'F26', 'F3': 'F1', 'F18': 'F9', 'F17': 'F3', 'F20': 'F15', 'F16': 'F24', 'F30': 'F20', 'F10': 'F22', 'F6': 'F27', 'F14': 'F23', 'F15': 'F6', 'F25': 'F13', 'F13': 'F33', 'F32': 'F10', 'F31': 'F4', 'F29': 'F19', 'F4': 'F2', 'F9': 'F12', 'F26': 'F25', 'F8': 'F7', 'F11': 'F28', 'F19': 'F21', 'F12': 'F8', 'F2': 'F18', 'F1': 'F5'}

{'C1': 'C2', 'C2': 'C1'}

Fraud

{'C2': 'Not Fraud', 'C1': 'Fraud'}

GradientBoostingClassifier

C2

Paris House Classification

Because the prediction probability of C1 is barely 0.70 percent, the classifier outputs the label C2 with near 100 percent confidence based on the values of the input attributes. The effects of F11, F10, and F16 on the aforementioned classification decision are significant. The values of these features are given greater emphasis by the classifier than the others. F16 is has a negative impact among these top features, pushing the prediction judgement towards the least likely class, C1 whereas on the other hand, F11 and F10 are referred to as positive features since they improve the likelihood of the C2 label rather than the C1 label. Finally, unlike the others, the values of F1, F13, F6, and F3 have only a little influence on the label selection made here.

{'C2': '99.30%', 'C1': '0.70%'}

{'F11': 'isNewBuilt', 'F16': 'hasYard', 'F10': 'hasPool', 'F7': 'hasStormProtector', 'F5': 'made', 'F9': 'hasGuestRoom', 'F15': 'squareMeters', 'F12': 'floors', 'F2': 'cityCode', 'F4': 'basement', 'F17': 'price', 'F14': 'numPrevOwners', 'F8': 'numberOfRooms', 'F1': 'attic', 'F13': 'cityPartRange', 'F6': 'garage', 'F3': 'hasStorageRoom'}

{'F3': 'F11', 'F1': 'F16', 'F2': 'F10', 'F4': 'F7', 'F12': 'F5', 'F16': 'F9', 'F6': 'F15', 'F8': 'F12', 'F9': 'F2', 'F13': 'F4', 'F17': 'F17', 'F11': 'F14', 'F7': 'F8', 'F14': 'F1', 'F10': 'F13', 'F15': 'F6', 'F5': 'F3'}

{'C1': 'C2', 'C2': 'C1'}

Basic

{'C2': 'Basic', 'C1': 'Luxury'}

SGDClassifier

C3

Flight Price-Range Classification

The classification algorithm arrived at the prediction output based on the variables or information supplied about the case under consideration. The prediction probabilities across the three-class labels, C2, C3, and C1, respectively, are 28.17%, 50.21%, and 21.62%, making C3 the label assigned by the algorithm, judged based on the prediction probabilities. The attributions analysis suggests that F2, F6, F3, and F5 are the positive features that increase the algorithm's prediction response in favour of C3. On the other hand, F8, F12, F1, F4, F7, F11, F9, and F10 have negative contributions in support of labelling the case as either C2 or C1. Overall, judging by the degree of contributions of the positive features, it is not surprising that the algorithm is moderately certain that neither C2 nor C1 is the most probable label for the case under consideration here.

{'C2': '28.17%', 'C3': '50.21%', 'C1': '21.62%'}

{'F2': 'Airline', 'F6': 'Total_Stops', 'F3': 'Arrival_minute', 'F8': 'Journey_day', 'F5': 'Dep_hour', 'F12': 'Source', 'F1': 'Dep_minute', 'F4': 'Duration_hours', 'F7': 'Destination', 'F11': 'Journey_month', 'F9': 'Duration_mins', 'F10': 'Arrival_hour'}

{'F9': 'F2', 'F12': 'F6', 'F6': 'F3', 'F1': 'F8', 'F3': 'F5', 'F10': 'F12', 'F4': 'F1', 'F7': 'F4', 'F11': 'F7', 'F2': 'F11', 'F8': 'F9', 'F5': 'F10'}

{'C2': 'C2', 'C3': 'C3', 'C1': 'C1'}

Moderate

{'C2': 'Low', 'C3': 'Moderate', 'C1': 'High'}

RandomForestClassifier

C1

Paris House Classification

Judging based on the information provided on the case under consideration, the model outputs that the prediction probability of C2 is only 0.48%, indicating that with about 99.52% certainty, the true label here is C1 and in simple terms, the model is very confident that the true label for the case under consideration is C1. The higher degree of certainty in the above classification can be attributed solely to the positive contributions of influential features F1, F7, and F15. Analysis indicates that all the remaining features such as F3, F13, F16, F5, and F9 have moderate to low contributions towards the prediction conclusions above, whereas F8, F2, F11, and F4 are the least relevant features here. The very marginal decrease in the C1's prediction likelihood could be attributed to the influence of negative features F13, F5, F6, F4, and F2 since their contributions support labelling the case as C2 instead. Moderate positive features further driving the model to label this case as C1 are F3, F16, F12, and F9.

{'C1': '99.52%', 'C2': '0.48%'}

{'F1': 'isNewBuilt', 'F7': 'hasYard', 'F15': 'hasPool', 'F3': 'made', 'F13': 'hasStormProtector', 'F16': 'hasGuestRoom', 'F5': 'squareMeters', 'F9': 'floors', 'F12': 'price', 'F10': 'cityCode', 'F14': 'basement', 'F6': 'numPrevOwners', 'F17': 'cityPartRange', 'F8': 'numberOfRooms', 'F2': 'attic', 'F11': 'garage', 'F4': 'hasStorageRoom'}

{'F3': 'F1', 'F1': 'F7', 'F2': 'F15', 'F12': 'F3', 'F4': 'F13', 'F16': 'F16', 'F6': 'F5', 'F8': 'F9', 'F17': 'F12', 'F9': 'F10', 'F13': 'F14', 'F11': 'F6', 'F10': 'F17', 'F7': 'F8', 'F14': 'F2', 'F15': 'F11', 'F5': 'F4'}

{'C1': 'C1', 'C2': 'C2'}

Basic

{'C1': 'Basic', 'C2': 'Luxury'}

GradientBoostingClassifier

C2

Basketball Players Career Length Prediction

Judging based on the values of the variables passed to the model with respect to the case under consideration, the output labelling decision is as follows: there is about an 83.98% chance that C2 is the correct label, whereas the likelihood of C1 is only 16.02%, hence the label choice with a higher confidence level is C2. The top-variables influencing this decision are F15, F18, F12, and F16, while the least important variables are F6, F4, and F5. According to the variable contributions analysis performed, only the input variables F7, F8, F19, and F2 exhibit negative attributions, pushing the prediction decision towards the alternative label, C1. The other variables positively support the C2 prediction, shifting the verdict strongly away from the C1 class. In conclusion, positive variables such as F15, F18, F12, F16, F10, and F17 have a higher joint contribution compared to the negative features, which can explain why the model is certain that C2 is the most probable label.

{'C2': '83.98%', 'C1': '16.02%'}

{'F15': 'GamesPlayed', 'F18': 'OffensiveRebounds', 'F12': 'FreeThrowPercent', 'F16': 'FieldGoalPercent', 'F3': '3PointPercent', 'F7': '3PointAttempt', 'F17': 'FieldGoalsMade', 'F10': 'Blocks', 'F11': 'DefensiveRebounds', 'F14': 'Turnovers', 'F8': 'Rebounds', 'F19': 'MinutesPlayed', 'F1': 'FreeThrowAttempt', 'F13': 'Assists', 'F2': '3PointMade', 'F9': 'FieldGoalsAttempt', 'F6': 'PointsPerGame', 'F5': 'Steals', 'F4': 'FreeThrowMade'}

{'F1': 'F15', 'F13': 'F18', 'F12': 'F12', 'F6': 'F16', 'F9': 'F3', 'F8': 'F7', 'F4': 'F17', 'F18': 'F10', 'F14': 'F11', 'F19': 'F14', 'F15': 'F8', 'F2': 'F19', 'F11': 'F1', 'F16': 'F13', 'F7': 'F2', 'F5': 'F9', 'F3': 'F6', 'F17': 'F5', 'F10': 'F4'}

{'C2': 'C2', 'C1': 'C1'}

More than 5

{'C2': 'More than 5', 'C1': 'Less than 5'}

RandomForestClassifier

C2

Mobile Price-Range Classification

The model predicts the class label C2 for the given test instance with a likelihood of about 69.23%. However, there is about a 30.77% chance that the true class label is C1, while the others, C3 and C4, have a 0.0% likelihood. The top features contributing to this prediction decision are F1, F20, F18, and F12, whereas the least important are F11, F15, and F19. Among the top features, while F1 and F20 have values that shift the prediction decision towards the C2 class label, the values of F18 and F12 suggest that the true label could likely be C1. For the features with moderate influence on the decision, F4, F5, F17, and F2 have negative contributions, further decreasing the confidence level in the C2 assignment. On the other hand, the moderate positive influences of F8, F16, F6, F14, and F7 drive the decision further towards the C2 label. Considering the attributions of the input features, it is surprising that the confidence level is just 69.23% since the top feature, F1, has the highest contribution among all the input features. Finally, the values of F10, F15, and F19, though shown to be less important when deciding the correct label for the given case, have positive contributions to the prediction with respect to the given case.

{'C3': '0.00%', 'C2': '69.23%', 'C1': '30.77%', 'C4': '0.00%'}

{'F1': 'ram', 'F20': 'touch_screen', 'F18': 'int_memory', 'F12': 'battery_power', 'F8': 'mobile_wt', 'F16': 'sc_w', 'F4': 'four_g', 'F14': 'talk_time', 'F6': 'sc_h', 'F7': 'wifi', 'F5': 'fc', 'F17': 'three_g', 'F2': 'dual_sim', 'F9': 'n_cores', 'F13': 'px_height', 'F3': 'blue', 'F10': 'clock_speed', 'F11': 'px_width', 'F15': 'm_dep', 'F19': 'pc'}

{'F11': 'F1', 'F19': 'F20', 'F4': 'F18', 'F1': 'F12', 'F6': 'F8', 'F13': 'F16', 'F17': 'F4', 'F14': 'F14', 'F12': 'F6', 'F20': 'F7', 'F3': 'F5', 'F18': 'F17', 'F16': 'F2', 'F7': 'F9', 'F9': 'F13', 'F15': 'F3', 'F2': 'F10', 'F10': 'F11', 'F5': 'F15', 'F8': 'F19'}

{'C3': 'C3', 'C4': 'C2', 'C2': 'C1', 'C1': 'C4'}

r2

{'C3': 'r1', 'C2': 'r2', 'C1': 'r3', 'C4': 'r4'}

KNeighborsClassifier

C1

Water Quality Classification

The given case is likely C1 with a confidence level of 87.50% judged based on the values of the input features supplied to the classifier and according to the attributions analysis, F8 and F2 have a high degree of impact. F9, F6, F1, F5, and F3 have a moderate degree of impact while on the contrary F4 and F7 have little impact. Examining further, the values of F8, F2, F9, and F6 all have a positive influence on the classifier supporting the label assignment decision for the given test case. F1 and F3 are also positively supporting features, whereas F5 has a negative influence on the final classification. Finally, F4 and F7 both have very little contributions, though F7 has significantly less than even F4.

{'C1': '87.50%', 'C2': '12.50%'}

{'F8': 'Hardness', 'F2': 'Sulfate', 'F9': 'Solids', 'F6': 'ph', 'F1': 'Organic_carbon', 'F3': 'Conductivity', 'F5': 'Trihalomethanes', 'F4': 'Turbidity', 'F7': 'Chloramines'}

{'F2': 'F8', 'F5': 'F2', 'F3': 'F9', 'F1': 'F6', 'F7': 'F1', 'F6': 'F3', 'F8': 'F5', 'F9': 'F4', 'F4': 'F7'}

{'C2': 'C1', 'C1': 'C2'}

Not Portable

{'C1': 'Not Portable', 'C2': 'Portable'}

RandomForestClassifier

C3

Mobile Price-Range Classification

The label for this example is estimated to be C3 among the four possible classes, with a 73.08 percent chance of being true. C1 is the next most likely label, with a probability of roughly 26.92 percent. The above prediction assessment is mostly dependent on the values of the variables F14, F12, F5, F1, and F17. F14 had the greatest influence, followed by F5, F12, F17, and F1. The positive variables F14, F12, F16, and F10 outnumber the negative variables F5, F17, F1, and F18. Twelve of the twenty variables have values that tilt the prediction towards one of the three other probable classifications. As a result, it is not unexpected that the model is not completely certain of the C3 assigned. Given that the chance of C3's being accurate is 73.08 percent, the model appears to be relatively confident in its final judgement for the data instance under review.

{'C3': '73.08%', 'C1': '26.92%', 'C2': '0.00%', 'C4': '0.00%'}

{'F14': 'ram', 'F5': 'px_width', 'F12': 'battery_power', 'F17': 'px_height', 'F1': 'n_cores', 'F16': 'dual_sim', 'F10': 'touch_screen', 'F18': 'int_memory', 'F3': 'wifi', 'F15': 'fc', 'F7': 'four_g', 'F8': 'm_dep', 'F20': 'pc', 'F4': 'mobile_wt', 'F9': 'talk_time', 'F19': 'three_g', 'F11': 'sc_h', 'F13': 'sc_w', 'F6': 'blue', 'F2': 'clock_speed'}

{'F11': 'F14', 'F10': 'F5', 'F1': 'F12', 'F9': 'F17', 'F7': 'F1', 'F16': 'F16', 'F19': 'F10', 'F4': 'F18', 'F20': 'F3', 'F3': 'F15', 'F17': 'F7', 'F5': 'F8', 'F8': 'F20', 'F6': 'F4', 'F14': 'F9', 'F18': 'F19', 'F12': 'F11', 'F13': 'F13', 'F15': 'F6', 'F2': 'F2'}

{'C2': 'C3', 'C1': 'C1', 'C4': 'C2', 'C3': 'C4'}

r1

{'C3': 'r1', 'C1': 'r2', 'C2': 'r3', 'C4': 'r4'}

BernoulliNB

C1

Personal Loan Modelling

The model has classified the instance as C1 due to the effects of the following features: F6, F5, F3, and F7. Based on the values of these variables, the likelihood of the C1 label is 65.51 percent. F7 and F3 are the top positively contributing variables, whereas F6 and F5 are the most adversely contributing variables. Unlike F7 and F3, which have greater influences on the model's prediction choice in this situation, F2 and F8 have fairly modest positive influences. Finally, F1, F9, and F4 show negative predictive effects, however, as compared to F6, their attributions are modest.

{'C2': '34.49%', 'C1': '65.51%'}

{'F7': 'CD Account', 'F3': 'Income', 'F6': 'CCAvg', 'F5': 'Securities Account', 'F2': 'Education', 'F8': 'Mortgage', 'F1': 'Age', 'F9': 'Family', 'F4': 'Extra_service'}

{'F8': 'F7', 'F2': 'F3', 'F4': 'F6', 'F7': 'F5', 'F5': 'F2', 'F6': 'F8', 'F1': 'F1', 'F3': 'F9', 'F9': 'F4'}

{'C1': 'C2', 'C2': 'C1'}

Accept

{'C2': 'Reject', 'C1': 'Accept'}

DecisionTreeClassifier

C2

Insurance Churn

Considering the predicted likelihoods across the classes, C2 is confidently chosen as the true label since its likelihood is 93.27%, implying that the likelihood of C1 is only about 6.73%. F14 and F4 are the two features with a very strong positive influence, favouring the prediction of class C2. The following features have a moderate effect and are listed in descending order of influence: F15 and F2 have a negative effect, while F3 and F1 have a positive effect on the prediction of C2. Similar to F15 and F2, the features F10 and F11 also negatively affected the prediction decision. Finally, the values of F13, F9, F5, and F12 are the least important to the model decision for this case.

{'C1': '6.73%', 'C2': '93.27%'}

{'F14': 'feature15', 'F4': 'feature14', 'F15': 'feature10', 'F2': 'feature11', 'F3': 'feature5', 'F1': 'feature13', 'F10': 'feature4', 'F11': 'feature3', 'F6': 'feature12', 'F16': 'feature1', 'F7': 'feature7', 'F8': 'feature2', 'F13': 'feature6', 'F9': 'feature0', 'F5': 'feature9', 'F12': 'feature8'}

{'F9': 'F14', 'F8': 'F4', 'F4': 'F15', 'F5': 'F2', 'F15': 'F3', 'F7': 'F1', 'F14': 'F10', 'F13': 'F11', 'F6': 'F6', 'F11': 'F16', 'F1': 'F7', 'F12': 'F8', 'F16': 'F13', 'F10': 'F9', 'F3': 'F5', 'F2': 'F12'}

{'C1': 'C1', 'C2': 'C2'}

Leave

{'C1': 'Stay', 'C2': 'Leave'}

GradientBoostingClassifier

C1

Basketball Players Career Length Prediction

The classification output is C1, however, the classifier is somewhat unsure about this prediction decision because the corresponding predicted probability is only 55.19%. F14 is by far the most influential feature whereas F7, F17, and F5 have been recognised as having the biggest effect on prediction output here after F14. The combination of F14, F7, F17, F5, and F11 features has resulted in the classification choice being altered from C1 to C2. While F4, F10, and F18 all have a minor influence on the classification, F4 is the only one that has a positive impact on the C1 classification. In this case, many features had lower influence on the prediction, with F15, F19, F1, F6, and F2 having a marginal effect.

{'C2': '44.81%', 'C1': '55.19%'}

{'F14': 'GamesPlayed', 'F7': 'OffensiveRebounds', 'F17': 'FieldGoalPercent', 'F5': 'FreeThrowPercent', 'F11': '3PointPercent', 'F4': '3PointAttempt', 'F10': 'FieldGoalsMade', 'F18': 'Blocks', 'F13': 'DefensiveRebounds', 'F12': 'Turnovers', 'F9': 'Rebounds', 'F3': 'MinutesPlayed', 'F16': 'FreeThrowAttempt', 'F8': '3PointMade', 'F15': 'Assists', 'F19': 'PointsPerGame', 'F1': 'FreeThrowMade', 'F6': 'FieldGoalsAttempt', 'F2': 'Steals'}

{'F1': 'F14', 'F13': 'F7', 'F6': 'F17', 'F12': 'F5', 'F9': 'F11', 'F8': 'F4', 'F4': 'F10', 'F18': 'F18', 'F14': 'F13', 'F19': 'F12', 'F15': 'F9', 'F2': 'F3', 'F11': 'F16', 'F7': 'F8', 'F16': 'F15', 'F3': 'F19', 'F10': 'F1', 'F5': 'F6', 'F17': 'F2'}

{'C2': 'C2', 'C1': 'C1'}

Less than 5

{'C2': 'More than 5', 'C1': 'Less than 5'}

LogisticRegression

C1

Customer Churn Modelling

Judging based on the values of the input variables, the classification algorithm labels the case as C1 since its prediction likelihood is equal to 88.69%. The prediction decision is primarily based on the contributions of F2, F9, and F6, however, F5, F1, and F4 are shown to be the least important variables. Regarding the direction of influence of the variables, F2, F6, F8, F5, and F1 are the positive variables that increase the odds of C1 being the correct label. Driving the prediction toward the alternative label, C2, are the variables F9, F10, F7, F3, and F4. Owing to the fact that the most influential variables, F2 and F6, have strong positive attributions, outweighing the contributions of the negative variables, it is not surprising that the algorithm is certain about the decision made.

{'C1': '88.69%', 'C2': '11.31%'}

{'F2': 'IsActiveMember', 'F6': 'NumOfProducts', 'F9': 'Geography', 'F10': 'Gender', 'F7': 'Age', 'F3': 'CreditScore', 'F8': 'EstimatedSalary', 'F5': 'Balance', 'F1': 'Tenure', 'F4': 'HasCrCard'}

{'F9': 'F2', 'F7': 'F6', 'F2': 'F9', 'F3': 'F10', 'F4': 'F7', 'F1': 'F3', 'F10': 'F8', 'F6': 'F5', 'F5': 'F1', 'F8': 'F4'}

{'C2': 'C1', 'C1': 'C2'}

Stay

{'C1': 'Stay', 'C2': 'Leave'}

BernoulliNB

C1

Water Quality Classification

The classification algorithm predicts class C1 with a confidence level of 61.55% and this implies that the probability of the alternative label is only 38.45%. In this case, the top features driving the prediction decision are F4, F3, F1, and F9, followed by F6, F7, F5, F8, and finally F2. Based on the inspections performed to understand the direction of influence of the input features, it can be concluded that F4 has the strongest positive contribution, while F1 has the strongest negative contribution and conversely, all the remaining features have moderate contributions. The other positive features are F3, F6, F7, and F8, whereas the remaining negatives are F9, F5, and F2. All things considered, the influence of the negative features indicates that the likelihood of the C2 label is 38.45% while the positive contributions push the prediction higher towards C1 resulting in the 61.55% prediction confidence.

{'C1': '61.55%', 'C2': '38.45%'}

{'F4': 'Sulfate', 'F3': 'ph', 'F1': 'Trihalomethanes', 'F9': 'Chloramines', 'F6': 'Organic_carbon', 'F7': 'Hardness', 'F5': 'Solids', 'F8': 'Turbidity', 'F2': 'Conductivity'}

{'F5': 'F4', 'F1': 'F3', 'F8': 'F1', 'F4': 'F9', 'F7': 'F6', 'F2': 'F7', 'F3': 'F5', 'F9': 'F8', 'F6': 'F2'}

{'C1': 'C1', 'C2': 'C2'}

Not Portable

{'C1': 'Not Portable', 'C2': 'Portable'}

RandomForestClassifier

C1

Flight Price-Range Classification

The classification model's decision about the true label for the case is based on the information provided to it. Among the three labels, C1, C2, and C3, the model shows without a doubt that neither C2 nor C3 is the true label, given that the probability of C1 being the true label is 100.0%. F7, F9, and F4 are the main contributing factors or variables in the final verdict here since their respective influence outranks the remaining variables. In fact, analysis indicates that F8, F2, and F11 are the least influential variables since they receive little emphasis from the model when making the labelling decision here. In between F7, F9, and F4, and F8, F2, F5, and F11, are the variables such as F12, F1, F3, and F6 with moderate influence on the classification decision here. Among the variables passed to the model, only F12, F6, and F8 are shown to have negative contributions, which suggests that perhaps the true label could be either of the remaining labels. However, given the 100.0% predicted likelihood of C1, it is reasonable to deduce that the positive variables, such as F7, F9, F4, F1, F10, and F3, significantly influence the model's judgement towards C1.

{'C1': '100.00%', 'C2': '0.00%', 'C3': '0.00%'}

{'F7': 'Duration_hours', 'F9': 'Airline', 'F4': 'Total_Stops', 'F12': 'Journey_day', 'F1': 'Source', 'F3': 'Duration_mins', 'F6': 'Arrival_hour', 'F10': 'Destination', 'F5': 'Arrival_minute', 'F8': 'Dep_minute', 'F2': 'Journey_month', 'F11': 'Dep_hour'}

{'F7': 'F7', 'F9': 'F9', 'F12': 'F4', 'F1': 'F12', 'F10': 'F1', 'F8': 'F3', 'F5': 'F6', 'F11': 'F10', 'F6': 'F5', 'F4': 'F8', 'F2': 'F2', 'F3': 'F11'}

{'C2': 'C1', 'C3': 'C2', 'C1': 'C3'}

Low

{'C1': 'Low', 'C2': 'Moderate', 'C3': 'High'}

LogisticRegression

C1

Basketball Players Career Length Prediction

According to the model, C1 is the class with the higher probability, which is equal to 52.57 percent, of being the label for this selected instance or case. Conversely, there is a 47.43 percent chance that C2 is the correct label showing that the model is less certain about the classification verdict in this case. This uncertainty can be linked to the fact that the majority of variables have values that favour assigning C2. The only variables increasing the model's response to prediction C1 are the positive variables namely: F8, F14, F18, F1, F2, F17, and F4. The top negative variables decreasing the likelihood of C1 are F19 and F13 supported by other negative variables, F7, F16, and F11, that further shift the verdict towards C2.

{'C2': '47.43%', 'C1': '52.57%'}

{'F19': '3PointMade', 'F8': '3PointAttempt', 'F13': 'FreeThrowMade', 'F14': 'FreeThrowAttempt', 'F7': 'GamesPlayed', 'F16': 'OffensiveRebounds', 'F11': 'FieldGoalsAttempt', 'F18': 'DefensiveRebounds', 'F10': 'Assists', 'F1': 'MinutesPlayed', 'F2': 'FieldGoalsMade', 'F12': 'Blocks', 'F9': 'Rebounds', 'F15': 'FieldGoalPercent', 'F17': 'Steals', 'F5': 'PointsPerGame', 'F6': 'FreeThrowPercent', 'F4': 'Turnovers', 'F3': '3PointPercent'}

{'F7': 'F19', 'F8': 'F8', 'F10': 'F13', 'F11': 'F14', 'F1': 'F7', 'F13': 'F16', 'F5': 'F11', 'F14': 'F18', 'F16': 'F10', 'F2': 'F1', 'F4': 'F2', 'F18': 'F12', 'F15': 'F9', 'F6': 'F15', 'F17': 'F17', 'F3': 'F5', 'F12': 'F6', 'F19': 'F4', 'F9': 'F3'}

{'C2': 'C2', 'C1': 'C1'}

Less than 5

{'C2': 'More than 5', 'C1': 'Less than 5'}

RandomForestClassifier

C2

Printer Sales

According to the predicted likelihoods across the classes, C1 has a 17.0% chance of being the true label for the given data or case, implying that C2 is the most likely label. F20, F8, and F25 are the most important factors that led to the classification judgments above. The remaining factors have a minor or non-existent impact on the classifier. The classifier most likely ignored the values of F10, F19, F11, F13, F12, and F9 when giving a label to this case since their relative degrees of impact are extremely near to zero. F21, F14, F5, F17, F23, and F1 are considered negative factors among the significant factors because their contributions to the choice tend to reduce the chance that C2 is the correct label. These negatives features lend themselves to the case being classified as C1 but the remaining features contribute positively, raising the likelihood of the C2 classification.

{'C2': '83.00%', 'C1': '17.00%'}

{'F20': 'X8', 'F8': 'X24', 'F25': 'X1', 'F16': 'X2', 'F15': 'X10', 'F6': 'X15', 'F21': 'X25', 'F26': 'X23', 'F4': 'X18', 'F14': 'X4', 'F5': 'X7', 'F22': 'X17', 'F3': 'X3', 'F2': 'X22', 'F17': 'X5', 'F23': 'X9', 'F1': 'X12', 'F18': 'X19', 'F24': 'X11', 'F7': 'X16', 'F10': 'X14', 'F19': 'X21', 'F11': 'X20', 'F13': 'X13', 'F12': 'X6', 'F9': 'X26'}

{'F8': 'F20', 'F24': 'F8', 'F1': 'F25', 'F2': 'F16', 'F10': 'F15', 'F15': 'F6', 'F25': 'F21', 'F23': 'F26', 'F18': 'F4', 'F4': 'F14', 'F7': 'F5', 'F17': 'F22', 'F3': 'F3', 'F22': 'F2', 'F5': 'F17', 'F9': 'F23', 'F12': 'F1', 'F19': 'F18', 'F11': 'F24', 'F16': 'F7', 'F14': 'F10', 'F21': 'F19', 'F20': 'F11', 'F13': 'F13', 'F6': 'F12', 'F26': 'F9'}

{'C2': 'C2', 'C1': 'C1'}

Less

{'C2': 'Less', 'C1': 'More'}

RandomForestClassifier

C2

Credit Risk Classification

According to the ML model, C2 is the most likely class label, and we can conclude that the model is quite confident about the decision given that the probability of having C1 as the correct label is only 7.0%. For the case under study, analysis indicates that F4, F5, F6, and F9 are essentially the negative set of features that push the forecast higher towards C1 instead of C2, while F8, F3, F7, and F11 increase the odds of the prediction being equal to C2. In general, the most relevant feature is F8, while F1 and F2 are the least relevant features, with marginal influence on the above classification verdict. In summary, given the very strong positive influence of F8 together with the moderate influence of the other positives, F3, F11, and F7, it is not strange that the model chose to label the case as C2 instead of C1.

{'C2': '93.00%', 'C1': '7.00%'}

{'F8': 'fea_4', 'F6': 'fea_10', 'F3': 'fea_8', 'F4': 'fea_7', 'F11': 'fea_2', 'F7': 'fea_3', 'F5': 'fea_5', 'F9': 'fea_1', 'F10': 'fea_9', 'F2': 'fea_6', 'F1': 'fea_11'}

{'F4': 'F8', 'F10': 'F6', 'F8': 'F3', 'F7': 'F4', 'F2': 'F11', 'F3': 'F7', 'F5': 'F5', 'F1': 'F9', 'F9': 'F10', 'F6': 'F2', 'F11': 'F1'}

{'C1': 'C2', 'C2': 'C1'}

Low

{'C2': 'Low', 'C1': 'High'}

MLPClassifier

C1

Annual Income Earnings

Because the confidence level associated with the other class, C2, is just 2.29%, the model predicts that the given example is likely C1 and to be specific, the model is quite certain that the right label for the given case is C1. All the features are shown to have some degree of influence on the decision above, with F6 and F7 being the least relevant features, while F13 and F14 are the top features. From the analysis performed to understand how each feature contributes to the above prediction assertion, only the features F5, F11, F2, F8, F3, and F7, have negative influences, shifting the prediction verdict towards C2. The remaining features all contribute positively, strongly shifting the prediction towards the assigned label which could explain the prediction confidence level associated with label C1. The most positive features are F14, F10, and F13 with stronger push in favour of the output label and they are supported by other positive features such as F9, F4, F12, and F1 have a moderate degree of influence.

{'C2': '2.29%', 'C1': '97.71%'}

{'F13': 'Capital Gain', 'F14': 'Marital Status', 'F5': 'Capital Loss', 'F10': 'Relationship', 'F11': 'Hours per week', 'F12': 'Education', 'F1': 'Country', 'F9': 'Age', 'F4': 'Occupation', 'F8': 'Sex', 'F2': 'Education-Num', 'F3': 'Workclass', 'F6': 'fnlwgt', 'F7': 'Race'}

{'F11': 'F13', 'F6': 'F14', 'F12': 'F5', 'F8': 'F10', 'F13': 'F11', 'F4': 'F12', 'F14': 'F1', 'F1': 'F9', 'F7': 'F4', 'F10': 'F8', 'F5': 'F2', 'F2': 'F3', 'F3': 'F6', 'F9': 'F7'}

{'C1': 'C2', 'C2': 'C1'}

Above 50K

{'C2': 'Under 50K', 'C1': 'Above 50K'}

KNNClassifier

C2

Car Acceptability Valuation

The classifier made the prediction here based on the information provided about the case under consideration, and according to the classifier, the prediction probabilities or likelihoods across the labels C2 and C1 are 100.0% and 0.0%, respectively. All the input features are shown to have different degrees of influence on the final decision here by the classifier. The most influential features are F5 and F6, with F4 and F1 ranked as the least contributing factors. The values of F3 and F2 suggest that perhaps the true label could be C1 since they are the negative features. However, considering the confidence in C2, it is valid to conclude that the joint influence or contribution to the classification of the negative features with respect to the given case is outmatched by the joint positive attribution of F5, F6, F4, and F1.

{'C2': '100.00%', 'C1': '0.00%'}

{'F5': 'persons', 'F6': 'safety', 'F2': 'lug_boot', 'F3': 'buying', 'F4': 'doors', 'F1': 'maint'}

{'F4': 'F5', 'F6': 'F6', 'F5': 'F2', 'F1': 'F3', 'F3': 'F4', 'F2': 'F1'}

{'C1': 'C2', 'C2': 'C1'}

Unacceptable

{'C2': 'Unacceptable', 'C1': 'Acceptable'}

LogisticRegression

C2

Real Estate Investment

For the selected case, the model assigns the label C2. The prediction probability distribution across the classes C1 and C2 is 2.40% and 97.60%, respectively. The most important features considered for this prediction are F11, F8, F6, and F17, while on the other hand, the least relevant features with little contributions to the decision based on the analysis are F9, F7, F1, and F14. The top positive features Increasing the likelihood of the prediction being made are F11, F8, and F17. Pushing the prediction towards the alternative class C1, the top negative features are F6, F12, and F19. F18, F3, F16, F20, and F4 are some of the features that have a moderate impact on the classification decision in this case.

{'C1': '2.40%', 'C2': '97.60%'}

{'F11': 'Feature7', 'F8': 'Feature4', 'F6': 'Feature2', 'F17': 'Feature14', 'F12': 'Feature15', 'F19': 'Feature8', 'F18': 'Feature20', 'F3': 'Feature1', 'F16': 'Feature17', 'F4': 'Feature3', 'F20': 'Feature16', 'F2': 'Feature18', 'F10': 'Feature10', 'F5': 'Feature5', 'F13': 'Feature6', 'F15': 'Feature12', 'F9': 'Feature19', 'F7': 'Feature13', 'F1': 'Feature9', 'F14': 'Feature11'}

{'F11': 'F11', 'F9': 'F8', 'F1': 'F6', 'F17': 'F17', 'F4': 'F12', 'F3': 'F19', 'F20': 'F18', 'F7': 'F3', 'F6': 'F16', 'F8': 'F4', 'F18': 'F20', 'F19': 'F2', 'F13': 'F10', 'F2': 'F5', 'F10': 'F13', 'F15': 'F15', 'F5': 'F9', 'F16': 'F7', 'F12': 'F1', 'F14': 'F14'}

{'C2': 'C1', 'C1': 'C2'}

Invest

{'C1': 'Ignore', 'C2': 'Invest'}

MLPClassifier

C1

Vehicle Insurance Claims

The given instance was labelled as C1 by the model based on the values of its features. The model is about 79.64% certain about this prediction decision, hence, there is a slight chance that the label could be C2. Among the different features, the ones with the most impact on the model are F11, F12, F18, F2, and F16. The most negative feature is F11, and it is significantly pushing the narrative toward the prediction of C2. From this, it is foreseeable that there is a chance that the true label could be C2 which is about 20.36%. The influence of F11 and F18 is somewhat counterbalanced by the values of the features F12, F2, and F16. Other attributes that shift the decision in favour of C2 are F22 and F31. F23 shifts the decision further in the direction of C1 and in addition, F3 supports the model's prediction while the values of F15 and F14 of the given test case contradict the model's decision, decreasing the likelihood of C1. Among the features not relevant to this prediction decision for this case are F32, F28, F30, and F21.

{'C1': '79.64%', 'C2': '20.36%'}

{'F11': 'incident_severity', 'F12': 'insured_hobbies', 'F18': 'insured_relationship', 'F2': 'umbrella_limit', 'F16': 'insured_education_level', 'F22': 'authorities_contacted', 'F31': 'incident_type', 'F23': 'policy_csl', 'F3': 'number_of_vehicles_involved', 'F15': 'capital-loss', 'F14': 'property_damage', 'F8': 'insured_occupation', 'F1': 'age', 'F13': 'incident_state', 'F33': 'insured_zip', 'F4': 'collision_type', 'F29': 'property_claim', 'F7': 'injury_claim', 'F27': 'capital-gains', 'F24': 'witnesses', 'F32': 'incident_city', 'F28': 'police_report_available', 'F21': 'months_as_customer', 'F30': 'auto_year', 'F26': 'insured_sex', 'F20': 'policy_state', 'F6': 'vehicle_claim', 'F9': 'total_claim_amount', 'F19': 'bodily_injuries', 'F25': 'incident_hour_of_the_day', 'F5': 'policy_annual_premium', 'F10': 'policy_deductable', 'F17': 'auto_make'}

{'F27': 'F11', 'F23': 'F12', 'F24': 'F18', 'F5': 'F2', 'F21': 'F16', 'F28': 'F22', 'F25': 'F31', 'F19': 'F23', 'F10': 'F3', 'F8': 'F15', 'F31': 'F14', 'F22': 'F8', 'F2': 'F1', 'F29': 'F13', 'F6': 'F33', 'F26': 'F4', 'F15': 'F29', 'F14': 'F7', 'F7': 'F27', 'F12': 'F24', 'F30': 'F32', 'F32': 'F28', 'F1': 'F21', 'F17': 'F30', 'F20': 'F26', 'F18': 'F20', 'F16': 'F6', 'F13': 'F9', 'F11': 'F19', 'F9': 'F25', 'F4': 'F5', 'F3': 'F10', 'F33': 'F17'}

{'C1': 'C1', 'C2': 'C2'}

Not Fraud

{'C1': 'Not Fraud', 'C2': 'Fraud'}

RandomForestClassifier

C1

Ethereum Fraud Detection

According to the classification algorithm, the best label for the given case is C1, because there is little to no chance that C2 is the correct label. Not all of the features are found to contribute to the label given here. The following significant features are ordered in order of their effect on the algorithm's output: F24, F25, F33, F6, F9, F3, F36, F22, F32, F28, F8, F5, F17, F23, F1, F18, F12, F30, F11, F19. F21, F38, and F26, on the other hand, are unimportant features since they have almost no influence. Among the most influential features F24, F25, F33, F6, and F9, F33 is considered the most negative, dragging the verdict in a different direction, while the others have positive contributions, increasing the possibility that C1 is correct in this case. F36 is recognised as a positive feature with modest effect, whereas F3 and F22 are identified as negative features. Given that the majority of the top five attributes have positive contributions, boosting the likelihood that C1 is the correct label, it is not unexpected that the algorithm is quite confident in the assigned label's accuracy.

{'C2': '0.00%', 'C1': '100.00%'}

{'F24': ' ERC20 total Ether sent contract', 'F25': ' ERC20 min val rec', 'F33': 'total transactions (including tnx to create contract', 'F6': ' ERC20 max val rec', 'F9': ' Total ERC20 tnxs', 'F3': ' ERC20 uniq rec addr', 'F36': 'min val sent', 'F22': 'Time Diff between first and last (Mins)', 'F32': 'Sent tnx', 'F28': 'Avg min between received tnx', 'F8': 'min value received', 'F5': ' ERC20 total ether sent', 'F17': 'avg val sent', 'F23': 'max val sent', 'F1': 'Avg min between sent tnx', 'F18': 'Received Tnx', 'F12': ' ERC20 uniq sent token name', 'F30': 'Unique Sent To Addresses', 'F11': ' ERC20 uniq rec token name', 'F19': ' ERC20 uniq rec contract addr', 'F21': 'total Ether sent', 'F38': 'Number of Created Contracts', 'F26': ' ERC20 avg val sent', 'F31': ' ERC20 max val sent', 'F2': ' ERC20 min val sent', 'F27': ' ERC20 avg val rec', 'F13': 'Unique Received From Addresses', 'F10': 'max value received ', 'F7': ' ERC20 uniq sent addr.1', 'F14': 'total ether sent contracts', 'F35': 'avg val received', 'F16': ' ERC20 uniq sent addr', 'F34': 'min value sent to contract', 'F4': 'max val sent to contract', 'F29': ' ERC20 total Ether received', 'F37': 'avg value sent to contract', 'F20': 'total ether balance', 'F15': 'total ether received'}

{'F26': 'F24', 'F31': 'F25', 'F18': 'F33', 'F32': 'F6', 'F23': 'F9', 'F28': 'F3', 'F12': 'F36', 'F3': 'F22', 'F4': 'F32', 'F2': 'F28', 'F9': 'F8', 'F25': 'F5', 'F14': 'F17', 'F13': 'F23', 'F1': 'F1', 'F5': 'F18', 'F37': 'F12', 'F8': 'F30', 'F38': 'F11', 'F30': 'F19', 'F19': 'F21', 'F6': 'F38', 'F36': 'F26', 'F35': 'F31', 'F34': 'F2', 'F33': 'F27', 'F7': 'F13', 'F10': 'F10', 'F29': 'F7', 'F21': 'F14', 'F11': 'F35', 'F27': 'F16', 'F15': 'F34', 'F16': 'F4', 'F24': 'F29', 'F17': 'F37', 'F22': 'F20', 'F20': 'F15'}

{'C1': 'C2', 'C2': 'C1'}

Fraud

{'C2': 'Not Fraud', 'C1': 'Fraud'}

BernoulliNB

C1

Hotel Satisfaction

The classifier labbelled the given case as C1 with a confidence level of 98.89%, implying that the chance of C2 being the correct label is only about 1.11%. The classification output decision is solely based on the information supplied to the classifier about the case under review. We can rank the contributions of the features as follows: F8, F1, F11, F5, F2, F10, F4, F3, F15, F9, F14, F12, F7, F6, and F13. Among the top features, F8 is the only negative feature, increasing the probability of predicting the alternative label, C2. Other top features that are shifting the prediction towards C1 are F1, F11, and F5. Similar to F8, the features F4, F6, and F9 have negative contributions, supporting the generation of C2. By comparing the strong joint positive attribution to the joint negative attribution, it is evident why the classifier is very certain that C1 is the right label for this instance.

{'C1': '98.89%', 'C2': '1.11%'}

{'F8': 'Type of Travel', 'F1': 'Type Of Booking', 'F11': 'Common Room entertainment', 'F5': 'Stay comfort', 'F2': 'Cleanliness', 'F10': 'Hotel wifi service', 'F4': 'Other service', 'F3': 'Ease of Online booking', 'F15': 'Age', 'F9': 'Checkin\\/Checkout service', 'F14': 'Food and drink', 'F12': 'Departure\\/Arrival convenience', 'F7': 'purpose_of_travel', 'F6': 'Hotel location', 'F13': 'Gender'}

{'F3': 'F8', 'F4': 'F1', 'F12': 'F11', 'F11': 'F5', 'F15': 'F2', 'F6': 'F10', 'F14': 'F4', 'F8': 'F3', 'F5': 'F15', 'F13': 'F9', 'F10': 'F14', 'F7': 'F12', 'F2': 'F7', 'F9': 'F6', 'F1': 'F13'}

{'C2': 'C1', 'C1': 'C2'}

dissatisfied

{'C1': 'dissatisfied', 'C2': 'satisfied'}

RandomForestClassifier

C2

Used Cars Price-Range Prediction

The prediction probability associated with class C1 is 10.50%, while that of class C2 is 89.50%, therefore, it can be concluded that C2 is the most probable label for the given case according to the model. All the input features are shown to contribute to the above decision, and the ones with the strongest influence on the classification decision are F1, F10, and F8, but F9, F7, F5, and F3 are shown to be the least relevant features . Finally, the degree of influence of F4, F2, and F6 can be described as moderate. The model's high confidence can be attributed to the strong positive contributions of F10 and F1 which are supported by the contributions of the remaining positive features F4, F9, and F7. Conversely, shifting the prediction in favour of C1, the negative features F8, F2, F5, F6, and F3.

{'C1': '10.50%', 'C2': '89.50%'}

{'F10': 'Power', 'F1': 'car_age', 'F8': 'Transmission', 'F4': 'Fuel_Type', 'F2': 'Name', 'F6': 'Mileage', 'F9': 'Engine', 'F5': 'Owner_Type', 'F7': 'Kilometers_Driven', 'F3': 'Seats'}

{'F4': 'F10', 'F5': 'F1', 'F8': 'F8', 'F7': 'F4', 'F6': 'F2', 'F2': 'F6', 'F3': 'F9', 'F9': 'F5', 'F1': 'F7', 'F10': 'F3'}

{'C1': 'C1', 'C2': 'C2'}

High

{'C1': 'Low', 'C2': 'High'}

SVC

C1

Food Ordering Customer Churn Prediction

The model labels the case as C1 with fairly high confidence equal to 89.73%, whereas the likelihood of C2 is only 10.27%. Analysis shows that only 20 of the 46 input variables contribute to the prediction assertion above. The prediction judgement C1 is mainly based on the variables F28, F11, F14, and F12. F15, F22, F18, F40, F25, and F46 also contribute to the decision, however, their degree of influence is only moderate. According to the direction of influence analysis, F28, F12, F25, and F40 positively support the decision of the model to assign the label C1. However, F11, F22, F46, F14, F15, and F18 reduce the likelihood or chance that C1 is the true label for this particular test instance. The main variables with less influence on the above classification decision are F6, F4, F44, and F19.

{'C1': '89.73%', 'C2': '10.27%'}

{'F28': 'Ease and convenient', 'F11': 'Unaffordable', 'F12': 'Good Food quality', 'F14': 'Wrong order delivered', 'F15': 'Delay of delivery person picking up food', 'F22': 'Politeness', 'F18': 'Self Cooking', 'F46': 'Late Delivery', 'F40': 'Health Concern', 'F25': 'More Offers and Discount', 'F24': 'Easy Payment option', 'F35': 'Time saving', 'F3': 'Perference(P2)', 'F20': 'Gender', 'F27': 'Good Road Condition', 'F9': 'Google Maps Accuracy', 'F43': 'Good Taste ', 'F31': 'Good Tracking system', 'F1': 'Bad past experience', 'F16': 'Marital Status', 'F6': 'Influence of rating', 'F44': 'Delivery person ability', 'F4': 'Low quantity low time', 'F19': 'Age', 'F45': 'Less Delivery time', 'F17': 'High Quality of package', 'F26': 'Maximum wait time', 'F29': 'Number of calls', 'F34': 'Freshness ', 'F7': 'Temperature', 'F30': 'Residence in busy location', 'F41': 'Long delivery time', 'F23': 'Order Time', 'F5': 'Influence of time', 'F37': 'Order placed by mistake', 'F8': 'Missing item', 'F2': 'Delay of delivery person getting assigned', 'F36': 'Family size', 'F33': 'Unavailability', 'F32': 'Poor Hygiene', 'F39': 'More restaurant choices', 'F38': 'Perference(P1)', 'F10': 'Educational Qualifications', 'F21': 'Monthly Income', 'F13': 'Occupation', 'F42': 'Good Quantity'}

{'F10': 'F28', 'F23': 'F11', 'F15': 'F12', 'F27': 'F14', 'F26': 'F15', 'F42': 'F22', 'F17': 'F18', 'F19': 'F46', 'F18': 'F40', 'F14': 'F25', 'F13': 'F24', 'F11': 'F35', 'F9': 'F3', 'F2': 'F20', 'F35': 'F27', 'F34': 'F9', 'F45': 'F43', 'F16': 'F31', 'F21': 'F1', 'F3': 'F16', 'F38': 'F6', 'F37': 'F44', 'F36': 'F4', 'F1': 'F19', 'F39': 'F45', 'F40': 'F17', 'F32': 'F26', 'F41': 'F29', 'F43': 'F34', 'F44': 'F7', 'F33': 'F30', 'F24': 'F41', 'F31': 'F23', 'F30': 'F5', 'F29': 'F37', 'F28': 'F8', 'F25': 'F2', 'F7': 'F36', 'F22': 'F33', 'F20': 'F32', 'F12': 'F39', 'F8': 'F38', 'F6': 'F10', 'F5': 'F21', 'F4': 'F13', 'F46': 'F42'}

{'C2': 'C1', 'C1': 'C2'}

Return

{'C1': 'Return', 'C2': 'Go Away'}

MLPClassifier

C2

Annual Income Earnings

The label predicted for this case is C2 with very high confidence of approximately 97.71% which insinuates that there is a marginal possibility that C1 could be the label. The above classification decision is largely due to the values of F9, F4, F13, and F14. On the other hand, F8 and F12 are less relevant when the model is deciding the correct label for the case here. Digging deeper revealed that each feature either positively or negatively contribute to the prediction made here. Six features contradicted the classification decision, while the remaining ones positively supported the C2 prediction. The negative features driving the prediction towards C1 are F13, F1, F11, F5, F3, and F12 and countering their influence are the top positive features are F9, F4, F10, and F14.

{'C1': '2.29%', 'C2': '97.71%'}

{'F9': 'Capital Gain', 'F4': 'Marital Status', 'F13': 'Capital Loss', 'F14': 'Relationship', 'F1': 'Hours per week', 'F10': 'Education', 'F7': 'Country', 'F6': 'Age', 'F2': 'Occupation', 'F11': 'Sex', 'F5': 'Education-Num', 'F3': 'Workclass', 'F8': 'fnlwgt', 'F12': 'Race'}

{'F11': 'F9', 'F6': 'F4', 'F12': 'F13', 'F8': 'F14', 'F13': 'F1', 'F4': 'F10', 'F14': 'F7', 'F1': 'F6', 'F7': 'F2', 'F10': 'F11', 'F5': 'F5', 'F2': 'F3', 'F3': 'F8', 'F9': 'F12'}

{'C2': 'C1', 'C1': 'C2'}

Above 50K

{'C1': 'Under 50K', 'C2': 'Above 50K'}

SVM_linear

C1

Wine Quality Prediction

The likelihood of C1 being the correct label for the selected case or instance is 67.54% according to the classifier. This means, there is a 32.46% chance that C2 could be the label and the classification assertion above is influenced mainly by the variables F4, F3, F6, and F5. On the contrary, F9, F7, and F10 are deemed less important when deciding the correct label for this given case. Decreasing the likelihood of the predicted label , C1, are the variables F5, F11, F7, and F10, therefore, these negative variables support the alternative class C2. However, the collective or joint attribution of the top positive variables, F3, F4, and F6 is strong enough to tilt the classification in favour of C1.

{'C2': '32.46%', 'C1': '67.54%'}

{'F4': 'residual sugar', 'F3': 'volatile acidity', 'F6': 'alcohol', 'F5': 'fixed acidity', 'F8': 'chlorides', 'F11': 'sulphates', 'F2': 'citric acid', 'F1': 'free sulfur dioxide', 'F9': 'density', 'F7': 'total sulfur dioxide', 'F10': 'pH'}

{'F4': 'F4', 'F2': 'F3', 'F11': 'F6', 'F1': 'F5', 'F5': 'F8', 'F10': 'F11', 'F3': 'F2', 'F6': 'F1', 'F8': 'F9', 'F7': 'F7', 'F9': 'F10'}

{'C1': 'C2', 'C2': 'C1'}

high quality

{'C2': 'low_quality', 'C1': 'high quality'}

KNeighborsClassifier

C1

Credit Risk Classification

The confidence level score with respect to each class label suggests that this case should be labelled as C1. Specifically, there is about an 80.0% chance that C1 is the correct label. However, this implies that there is also about a 20.0% chance that it should be C2. The above prediction decision is based predominantly on the influence of the following features: F1, F6, F3, F5, F8, F11, and F10. According to the analysis, the features F1, F6, and F3 have a very strong positive influence, swinging the prediction decision towards C1. In contrast, the value of F5 also suggests the decision should be the alternative class, C2. Similar to F5, the values of F4, F8, and F11 indicate the label could be C2. However, the influence of these features is very small compared to F1, F6, F3, and F5. Finally, the attributes with a moderately low influence on the final prediction decision for this case include F10, F2, F7, and F9. The values of F10 and F9 have a negative attribution, while F2 and F7 have positive attributions.

{'C1': '80.00%', 'C2': '20.00%'}

{'F1': 'fea_4', 'F6': 'fea_8', 'F3': 'fea_2', 'F5': 'fea_9', 'F4': 'fea_6', 'F8': 'fea_10', 'F11': 'fea_1', 'F10': 'fea_11', 'F2': 'fea_7', 'F9': 'fea_3', 'F7': 'fea_5'}

{'F4': 'F1', 'F8': 'F6', 'F2': 'F3', 'F9': 'F5', 'F6': 'F4', 'F10': 'F8', 'F1': 'F11', 'F11': 'F10', 'F7': 'F2', 'F3': 'F9', 'F5': 'F7'}

{'C2': 'C1', 'C1': 'C2'}

Low

{'C1': 'Low', 'C2': 'High'}

BernoulliNB

C2

Job Change of Data Scientists

The classification algorithm is pretty confident that the correct label for the data under consideration is C2, wowever, it is noteworthy to consider that C1 has about a 15.13% chance of being the correct label. The predicted probability of each label is assigned based on the influence of features such as F8, F10, F2, and F4. However, the analysis shows that the values of F5, F3, F9, and F1 are less relevant when classifying the data. Only the features F2, F11, F5, F3, F9, and F1 have negative attributions, decreasing the predicted probability of the assigned label and one can say these features are shifting the prediction decision towards the label C1.

{'C1': '15.13%', 'C2': '84.87%'}

{'F8': 'city', 'F10': 'enrolled_university', 'F2': 'relevent_experience', 'F4': 'city_development_index', 'F11': 'experience', 'F6': 'education_level', 'F12': 'major_discipline', 'F7': 'last_new_job', 'F5': 'gender', 'F3': 'company_size', 'F9': 'company_type', 'F1': 'training_hours'}

{'F3': 'F8', 'F6': 'F10', 'F5': 'F2', 'F1': 'F4', 'F9': 'F11', 'F7': 'F6', 'F8': 'F12', 'F12': 'F7', 'F4': 'F5', 'F10': 'F3', 'F11': 'F9', 'F2': 'F1'}

{'C2': 'C1', 'C1': 'C2'}

Leave

{'C1': 'Stay', 'C2': 'Leave'}

GradientBoostingClassifier

C2

Printer Sales

The case, despite having features with considerable negative impact, also has numerous and measurable positive features, so the assignment of the label C2 by the model is very likely since the predicted probability is 91.95% which is very higher than that of C1. The F13, F23, and F10 were the most important features driving the model to arrive at the labelling assignment of C2. F3 and F18 have nearly identical positive attributions, while F24 and F1 has negative impacts, swinging the prediction towards a different label. However, the joint positive impact of F3, F13, F10, and F18 stands out over the impact of F23, F20, F1, and F24, favouring the prediction of the C2 model. All things considered, there are more features with a positive impact than those with negative impact; the mean attribution of the positive attributes is much larger which somewhat explains why the confidence level is very high. Above all, it is important to note that the prediction is made with less emphasis on the values of F21, F5, F8, and F26 hence they are practically irrelevant when it comes to labelling this case.

{'C1': '8.05%', 'C2': '91.95%'}

{'F13': 'X24', 'F23': 'X8', 'F10': 'X1', 'F20': 'X21', 'F1': 'X4', 'F18': 'X6', 'F3': 'X3', 'F24': 'X22', 'F12': 'X7', 'F2': 'X15', 'F6': 'X20', 'F4': 'X11', 'F14': 'X10', 'F16': 'X19', 'F7': 'X5', 'F19': 'X16', 'F15': 'X23', 'F22': 'X9', 'F9': 'X17', 'F11': 'X18', 'F5': 'X25', 'F25': 'X14', 'F26': 'X2', 'F8': 'X13', 'F17': 'X12', 'F21': 'X26'}

{'F24': 'F13', 'F8': 'F23', 'F1': 'F10', 'F21': 'F20', 'F4': 'F1', 'F6': 'F18', 'F3': 'F3', 'F22': 'F24', 'F7': 'F12', 'F15': 'F2', 'F20': 'F6', 'F11': 'F4', 'F10': 'F14', 'F19': 'F16', 'F5': 'F7', 'F16': 'F19', 'F23': 'F15', 'F9': 'F22', 'F17': 'F9', 'F18': 'F11', 'F25': 'F5', 'F14': 'F25', 'F2': 'F26', 'F13': 'F8', 'F12': 'F17', 'F26': 'F21'}

{'C1': 'C1', 'C2': 'C2'}

More

{'C1': 'Less', 'C2': 'More'}

LogisticRegression

C3

Cab Surge Pricing System

The predicted label is C3 given the predictability of C1 is 28.96% and that of C2 is 23.41%. Considering the probabilities of the classes, the model can be described as being moderately confident. The prediction of C3 can be attributed to the varying degree of contributions of the input features. Attribution analysis indicates that F9, F4, and F8 are considered the most influential. Those with moderate influence are F2, F3, F7, F11, F1, and F10, whereas on the contrary, the least influential ones are F5, F6, and F12. The analysis also revealed that not all the features contribute positively to the prediction decision and amongst the input features, the ones with negative attributions decreasing the likelihood of the C3 prediction are F4, F8, F7, F11, and F1 whereas conversely, the top positive features are F9, F2, and F3.

{'C1': '28.96%', 'C2': '23.41%', 'C3': '47.63%'}

{'F9': 'Type_of_Cab', 'F4': 'Confidence_Life_Style_Index', 'F8': 'Destination_Type', 'F2': 'Trip_Distance', 'F3': 'Cancellation_Last_1Month', 'F7': 'Life_Style_Index', 'F11': 'Customer_Rating', 'F1': 'Var3', 'F10': 'Var1', 'F5': 'Customer_Since_Months', 'F6': 'Var2', 'F12': 'Gender'}

{'F2': 'F9', 'F5': 'F4', 'F6': 'F8', 'F1': 'F2', 'F8': 'F3', 'F4': 'F7', 'F7': 'F11', 'F11': 'F1', 'F9': 'F10', 'F3': 'F5', 'F10': 'F6', 'F12': 'F12'}

{'C3': 'C1', 'C1': 'C2', 'C2': 'C3'}

C3

{'C1': 'Low', 'C2': 'Medium', 'C3': 'High'}

RandomForestClassifier

C1

Company Bankruptcy Prediction

The model outputs a predicted probability of 2.55% for the C2 label and 97.45% for the C1 label. Judging from above, the most probable class is C1. Hence, C1 is the assigned label by the model, with a very high confidence level. The top features contributing to the prediction assessment above are F50, F66, F59, F61, and F16. However, the values of about twenty features are deemed relevant while the remaining are regarded as irrelevant when classifying the given case. These irrelevant features include F68, F48, F71, and F93. Among the relevant features, F59, F1, F49, F5, F67, and F51 are shown to be the only positive features that increase the model's response in favour of the assigned label C1. In contrast, the majority of the relevant features, mainly F50, F66, F61, and F16, have negative contributions, decreasing the odds of the label C1, hence supporting the assignment of C2 to the given case.

{'C2': '2.55%', 'C1': '97.45%'}

{'F50': " Net Income to Stockholder's Equity", 'F66': ' Total income\\/Total expense', 'F59': ' Borrowing dependency', 'F61': ' Continuous interest rate (after tax)', 'F16': ' Net Value Per Share (B)', 'F87': ' Cash\\/Current Liability', 'F20': ' Net worth\\/Assets', 'F1': ' Fixed Assets Turnover Frequency', 'F49': ' Interest-bearing debt interest rate', 'F5': ' No-credit Interval', 'F35': ' Net Value Per Share (A)', 'F42': ' Long-term fund suitability ratio (A)', 'F67': ' Equity to Long-term Liability', 'F63': ' Realized Sales Gross Margin', 'F51': ' Current Asset Turnover Rate', 'F10': ' Working Capital to Total Assets', 'F47': ' Interest Coverage Ratio (Interest expense to EBIT)', 'F2': ' Working capitcal Turnover Rate', 'F32': ' Inventory Turnover Rate (times)', 'F36': ' After-tax net Interest Rate', 'F68': ' Working Capital\\/Equity', 'F48': ' Liability to Equity', 'F71': ' Operating Gross Margin', 'F93': ' Cash Flow Per Share', 'F24': ' Contingent liabilities\\/Net worth', 'F56': ' Operating Profit Per Share (Yuan ¥)', 'F11': ' Operating Profit Rate', 'F80': ' Net Worth Turnover Rate (times)', 'F8': ' Continuous Net Profit Growth Rate', 'F41': ' Long-term Liability to Current Assets', 'F91': ' Fixed Assets to Assets', 'F92': ' Inventory and accounts receivable\\/Net value', 'F39': ' Regular Net Profit Growth Rate', 'F65': ' Current Liability to Equity', 'F83': ' Equity to Liability', 'F29': ' Current Liability to Liability', 'F85': ' Operating profit\\/Paid-in capital', 'F72': ' Net Value Per Share (C)', 'F77': ' Operating Funds to Liability', 'F75': ' Current Liability to Current Assets', 'F21': ' Current Ratio', 'F27': ' Quick Assets\\/Current Liability', 'F82': ' Tax rate (A)', 'F13': ' After-tax Net Profit Growth Rate', 'F31': ' Per Share Net profit before tax (Yuan ¥)', 'F89': ' Total Asset Turnover', 'F25': ' CFO to Assets', 'F18': ' Cash Reinvestment %', 'F12': ' Net profit before tax\\/Paid-in capital', 'F62': ' Cash Flow to Equity', 'F37': ' Debt ratio %', 'F4': ' Current Liabilities\\/Liability', 'F55': ' Interest Expense Ratio', 'F76': ' Cash Flow to Sales', 'F26': ' Total Asset Growth Rate', 'F38': ' Inventory\\/Current Liability', 'F53': ' Allocation rate per person', 'F30': ' Operating Expense Rate', 'F33': ' Operating profit per person', 'F17': ' Net Income to Total Assets', 'F74': ' Net Value Growth Rate', 'F7': ' ROA(B) before interest and depreciation after tax', 'F45': ' Cash Flow to Liability', 'F70': ' Inventory\\/Working Capital', 'F64': ' Retained Earnings to Total Assets', 'F58': ' Total assets to GNP price', 'F43': ' Persistent EPS in the Last Four Seasons', 'F86': ' Total debt\\/Total net worth', 'F34': ' Quick Ratio', 'F23': ' Revenue per person', 'F84': ' Non-industry income and expenditure\\/revenue', 'F44': ' Cash\\/Total Assets', 'F78': ' ROA(A) before interest and % after tax', 'F57': ' ROA(C) before interest and depreciation before interest', 'F90': ' Research and development expense rate', 'F73': ' Cash Flow to Total Assets', 'F19': ' Pre-tax net Interest Rate', 'F40': ' Accounts Receivable Turnover', 'F81': ' Current Liability to Assets', 'F46': ' Quick Assets\\/Total Assets', 'F6': ' Total expense\\/Assets', 'F22': ' Operating Profit Growth Rate', 'F3': ' Average Collection Days', 'F14': ' Current Assets\\/Total Assets', 'F88': ' Current Liabilities\\/Equity', 'F54': ' Realized Sales Gross Profit Growth Rate', 'F52': ' Cash flow rate', 'F9': ' Total Asset Return Growth Rate Ratio', 'F15': ' Degree of Financial Leverage (DFL)', 'F28': ' Cash Turnover Rate', 'F79': ' Quick Asset Turnover Rate', 'F69': ' Revenue Per Share (Yuan ¥)', 'F60': ' Gross Profit to Sales'}

{'F59': 'F50', 'F57': 'F66', 'F3': 'F59', 'F12': 'F61', 'F27': 'F16', 'F32': 'F87', 'F84': 'F20', 'F22': 'F1', 'F1': 'F49', 'F56': 'F5', 'F42': 'F35', 'F52': 'F42', 'F23': 'F67', 'F83': 'F63', 'F61': 'F51', 'F67': 'F10', 'F60': 'F47', 'F73': 'F2', 'F18': 'F32', 'F79': 'F36', 'F68': 'F68', 'F66': 'F48', 'F62': 'F71', 'F65': 'F93', 'F64': 'F24', 'F63': 'F56', 'F58': 'F11', 'F55': 'F80', 'F54': 'F8', 'F69': 'F41', 'F74': 'F91', 'F70': 'F92', 'F85': 'F39', 'F92': 'F65', 'F91': 'F83', 'F90': 'F29', 'F89': 'F85', 'F88': 'F72', 'F87': 'F77', 'F86': 'F75', 'F82': 'F21', 'F71': 'F27', 'F81': 'F82', 'F80': 'F13', 'F78': 'F31', 'F77': 'F89', 'F76': 'F25', 'F75': 'F18', 'F72': 'F12', 'F53': 'F62', 'F47': 'F37', 'F51': 'F4', 'F14': 'F55', 'F25': 'F76', 'F24': 'F26', 'F21': 'F38', 'F20': 'F53', 'F19': 'F30', 'F17': 'F33', 'F16': 'F17', 'F15': 'F74', 'F13': 'F7', 'F50': 'F45', 'F11': 'F70', 'F10': 'F64', 'F9': 'F58', 'F8': 'F43', 'F7': 'F86', 'F6': 'F34', 'F5': 'F23', 'F4': 'F84', 'F26': 'F44', 'F28': 'F78', 'F29': 'F57', 'F30': 'F90', 'F49': 'F73', 'F48': 'F19', 'F2': 'F40', 'F46': 'F81', 'F45': 'F46', 'F44': 'F6', 'F43': 'F22', 'F41': 'F3', 'F40': 'F14', 'F39': 'F88', 'F38': 'F54', 'F37': 'F52', 'F36': 'F9', 'F35': 'F15', 'F34': 'F28', 'F33': 'F79', 'F31': 'F69', 'F93': 'F60'}

{'C1': 'C2', 'C2': 'C1'}

Yes

{'C2': 'No', 'C1': 'Yes'}

SVC

C1

Broadband Sevice Signup

The algorithm identifies the provided data or case as C1 with a greater level of certainty since the prediction probability of class C2 is just 0.07 percent as a result, C2 is less likely than C1. The influence of input features such as F5, F14, F18, F36, and F33 is mostly responsible for the classification verdict above with only F33 having a negative influence among them, slightly pulling the decision in favour of C2. F5, F14, F18, and F36, on the other hand, make considerable positive contributions in favour of assigning C1 to the data. F1, F40, F12, F24, F34, F23, F32, and F39 are some more features that have a modest effect on the algorithm's decision. But, not all features are demonstrated to influence the classification decision either negatively or positively to the aforementioned classification outcome and in reality, a number of these are demonstrated to be irrelevant for determining the suitable label for this case and these include F10, F19, F26, and F27. All in all, the most important features for this classification instance are F5 and F14, whereas F41 and F37 are the least important.

{'C1': '99.93%', 'C2': '0.07%'}

{'F5': 'X38', 'F14': 'X32', 'F18': 'X31', 'F36': 'X25', 'F33': 'X8', 'F1': 'X35', 'F40': 'X1', 'F32': 'X3', 'F24': 'X28', 'F39': 'X19', 'F34': 'X9', 'F12': 'X11', 'F23': 'X10', 'F8': 'X21', 'F35': 'X17', 'F9': 'X4', 'F42': 'X36', 'F22': 'X2', 'F20': 'X6', 'F11': 'X34', 'F19': 'X37', 'F10': 'X40', 'F26': 'X42', 'F27': 'X41', 'F2': 'X5', 'F28': 'X33', 'F29': 'X39', 'F16': 'X24', 'F30': 'X30', 'F3': 'X27', 'F4': 'X26', 'F38': 'X23', 'F15': 'X22', 'F25': 'X20', 'F21': 'X18', 'F13': 'X16', 'F31': 'X15', 'F6': 'X14', 'F37': 'X13', 'F41': 'X12', 'F17': 'X7', 'F7': 'X29'}

{'F35': 'F5', 'F29': 'F14', 'F28': 'F18', 'F23': 'F36', 'F6': 'F33', 'F32': 'F1', 'F40': 'F40', 'F2': 'F32', 'F26': 'F24', 'F17': 'F39', 'F7': 'F34', 'F9': 'F12', 'F8': 'F23', 'F19': 'F8', 'F15': 'F35', 'F3': 'F9', 'F33': 'F42', 'F1': 'F22', 'F4': 'F20', 'F31': 'F11', 'F34': 'F19', 'F37': 'F10', 'F38': 'F26', 'F39': 'F27', 'F41': 'F2', 'F30': 'F28', 'F36': 'F29', 'F22': 'F16', 'F27': 'F30', 'F25': 'F3', 'F24': 'F4', 'F21': 'F38', 'F20': 'F15', 'F18': 'F25', 'F16': 'F21', 'F14': 'F13', 'F13': 'F31', 'F12': 'F6', 'F11': 'F37', 'F10': 'F41', 'F5': 'F17', 'F42': 'F7'}

{'C1': 'C1', 'C2': 'C2'}

No

{'C1': 'No', 'C2': 'Yes'}

SVM

C3

Customer Churn Modelling

For the given dataset instance, the label assigned by the classifier is C3 since it has a predicted probability of about 89.16%. On the other hand, there is a 9.0% chance that C1 could be the appropriate label, whereas C2 only has a 1.84% chance of being the true label. The classifier arrived at this classification verdict chiefly due to the influence and contributions of variables such as F7, F2, F4, and F9. However, there is less emphasis on the values of F3, F8, and F10, since their impact on the classifier with respect to the given case is smaller compared to the other variables, hence they are the least ranked features. From the attribution analysis, there are four variables with negative contributions, pushing the verdict in the direction of C1. These negative variables are F7, F9, F6, and F5, and their influence on the classifier could explain why there is a little bit of doubt about the correctness of the C3 class assigned and the notable positive variables are F2, F1, F3, and F4.

{'C3': '89.16%', 'C1': '9.0%', 'C2': '1.84%'}

{'F7': 'IsActiveMember', 'F2': 'Age', 'F4': 'Geography', 'F9': 'NumOfProducts', 'F6': 'Gender', 'F1': 'Tenure', 'F5': 'CreditScore', 'F3': 'Balance', 'F8': 'EstimatedSalary', 'F10': 'HasCrCard'}

{'F9': 'F7', 'F4': 'F2', 'F2': 'F4', 'F7': 'F9', 'F3': 'F6', 'F5': 'F1', 'F1': 'F5', 'F6': 'F3', 'F10': 'F8', 'F8': 'F10'}

{'C1': 'C3', 'C2': 'C1', 'C3': 'C2'}

Stay

{'C3': 'Stay', 'C1': 'Leave', 'C2': 'Other'}