Datasets:

james-burton

/

aug-text-exps-v3

like 0

SVC

C2

Broadband Sevice Signup

With a higher level of certainty, the algorithm labels the given data or case as C2 because the predicted probability of class C2 is 99.93% while that of class C1 is only 0.07%. C1 is therefore less likely than C2 and the classification assertion or decision here is chiefly attributed to the impact of input features such as F6, F3, F36, F14, and F21. Among these relevant features, only F21 has a negative contribution, mildly dragging the verdict in favour of C1, whereas conversely, F6, F3, F36, and F14 have strong positive contributions in support of assigning C2 to the given data. Other features with moderate influence on the algorithm's verdict here include F23, F28, F30, F40, F32, F41, F4, and F15. However, some of the input features are shown to have negligible contribution to the abovementioned classification output and in fact, these include F33, F24, F29, and F9. In summary, the most vital features with respect to this classification instance are F6, F36, and F3 with positive contributions strongly increasing the algorithm's response towards label C2 hence the 99.93% predicted probability.

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

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

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

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

No

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

KNeighborsClassifier

C1

E-Commerce Shipping

There is uncertainty about the correct label for the given example since both labels, C1 and C2 are shown to have a 50.0% chance of being correct. The prediction decision above is mainly attributed to the influence of the input features F9, F10, and F8, while F1, F3, and F2 are deemed less important to the decision above. Looking at the direction of influence of each input feature, only F10, F10, F1, and F2 are shown to have a positive contribution, increasing the model's response towards assigning C1. All the remaining six features have a negative contribution towards the decision here, supporting the assignment of the other class.

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

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

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

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

On-time

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

DecisionTreeClassifier

C2

Credit Risk Classification

The classification model assigned the label C2 to the given example and given that the confidence level is 100.0%, we can be certain that the chances of C1 being the true label are negligible. The most relevant features controlling the prediction decision above are F7, F9, and F6. F11, F5, and F4 are among the least relevant features. Most of the properties have values that sway the decision towards the other C1 class. The only features that increase the odds that C2 is the correct label are F7, F2, and F5. It is strange that the model has 100.0% confidence in its prediction for the selected sample, given that only a small number of the input features contribute positively to reaching the C2 estimate.

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

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

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

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

High

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

LogisticRegression

C1

Hotel Satisfaction

The model prediction for the test case is C1 and the confidence level of this is almost 100%. From examining the contributions of variables or attributes, the values of F15 and F11 push the prediction verdict in favor of the other label. On the contrary, F1, F13, F7, and F6 have values with a positive influence that biases the classification decision towards label C1. While attributes F10 and F2 contradict the prediction made, F4 and F12 have values that support the model's prediction for the given case.

{'C1': '91.36%', 'C2': '8.64%'}

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

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

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

dissatisfied

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

SGDClassifier

C1

Company Bankruptcy Prediction

The label predicted by the classifier is C1 at a 71.80% confidence level. On the other hand, there is a 28.20% chance that C2 could be the label. The prediction can be mainly attributed to contributions from F50, F51, F36, and F63. Considerable positive contributions to the prediction here are from F50, F63, F52, and F51 since their values support the prediction of C1. Shifting the prediction towards C2 are the negative features F36, F11, F6, F43, and F77. There were some features with minuscule influence on prediction decision made for the case under consideration; these include F91, F70, and F8. In simple terms, the classifer deems the values of these features less important when assigning the label here.

{'C1': '71.80%', 'C2': '28.20%'}

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

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

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

No

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

BernoulliNB

C2

Employee Promotion Prediction

The model, making a classification decision based on the input variables, predicts the class C2 label for this case with a predicted likelihood equal to 54.21%. It also shows a 45.79% probability that C1 is the correct label. The classification decision made above is primarily influenced by the variables F6, F11, F4, F2, and F1. The three most influential variables, F6, F11, and F2, have a negative impact since their values are shifting the labelling decision in the direction of C1 instead of C2. Positive variables are F1, F4, F5, F3, and F9, supporting the model's class assignment decision for this situation and one can conclude that it is the influence of the positives that motivates the decision towards C2.

{'C2': '54.21%', 'C1': '45.79%'}

{'F6': 'KPIs_met >80%', 'F11': 'previous_year_rating', 'F2': 'avg_training_score', 'F1': 'department', 'F4': 'education', 'F5': 'recruitment_channel', 'F10': 'no_of_trainings', 'F7': 'length_of_service', 'F3': 'region', 'F8': 'age', 'F9': 'gender'}

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

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

Ignore

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

MLPClassifier

C2

Vehicle Insurance Claims

Based on the values of the input features, the classifier believes that the most probable label for the given data is C2, due to the fact that there is only a 19.30% chance that it could be C1 instead. The most influential features resulting in the decision or judgement above are F10, F12, F15, F31, F6, F16, and F8, though features such as F29, F19, F21, and F9 are indicated to have negligible contributions to the classification. Actually, the high certainty of the chosen label can be attributed to the very strong positive influence of F10 and the moderate positive influence of F12, F15, F17, and F31. Conversely, the negative features F16, F8, F6, and F33 reduce the likelihood of C2 since their values support labelling the case as C1.

{'C1': '19.30%', 'C2': '80.70%'}

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

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

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

Fraud

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

SGDClassifier

C1

House Price Classification

According to the classification algorithm with a very high confidence level, the correct label for the given data instance is C1. This prediction decision is heavily influenced by features such as F12, F4, F5, F10, F6, and F11. Among these top features, the only features with a negative contribution towards the assigned label are F11 and F6. With respect to the given instance, their negative contributions decrease the algorithm's response in favour of the least probable class. F8, F1, F13, F3, and F2 positively support the assignment of label C1. Conversely, F7 and F1 have a similar direction of influence as F6 and F11.

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

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

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

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

High

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

RandomForestClassifier

C2

Bike Sharing Demand

For the given data instance, the most probable class according to the classifier is C2 since the probability of C1 being the correct label is only about 10.0%. The most influential features resulting in the prediction decision above are F1, F10, and F3 which are shown to negatively contribute to the decision above since they strongly push the classifier towards assigning a different label. F11, F8, and F2 are shown to be the only features to positively contribute to the classification made here. Aside from the positive features, all the others negatively reduce the odds of the given data instance having C2 as its label.

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

{'F1': 'Functioning Day', 'F10': 'Rainfall(mm)', 'F3': 'Temperature', 'F11': 'Solar Radiation (MJ\\/m2)', 'F8': 'Seasons', 'F4': 'Wind speed (m\\/s)', 'F2': 'Holiday', 'F6': 'Visibility (10m)', 'F5': 'Dew point temperature', 'F9': 'Hour', 'F7': 'Snowfall (cm)', 'F12': 'Humidity(%)'}

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

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

Less than 500

{'C2': 'Less than 500', 'C1': 'More than 500'}

RandomForestClassifier

C1

Advertisement Prediction

Judging based on the information about the given case, the model outputs C1 with a prediction probability of 74.72%, however, it is vital to keep in mind that there is also a 25.28% probability that C2 could be the true label. The attribution analysis shows that all the input variables have varying degrees of influence on the model as it arrives at the abovementioned decision and the influence of the features can be ranked from the most relevant to the least relevant as follows: F6, F4, F2, F7, F3, F5, and F1. Across the input features, only F4 and F7 have negative attributions, reducing the likelihood of the predicted label which explain the 25.28% predicted likelihood of the C2 label. Therefore, F6, F2, F3, F5, and F1 are the positive input features pushing the decision higher towards C1 and away from C2.

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

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

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

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

Skip

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

DNN

C2

Credit Card Fraud Classification

The data is labelled C2 by the model as it has a somewhat greater prediction chance than C1. F13, F28, F23, F15, and F2 are the input variables that have the most impact on the above classification choice, whereas F9, F5, F12, F3, and F4 have the least influence. F13, F28, F2, and F15 are basically supporting the choice of the label in this scenario while on the contrary, F23, F7, and F19 are the primary negative factors. It's not unexpected that the model isn't 100 percent sure of the assigned label considering the degree of influence as well as the direction of influence of the variables.

{'C1': '48.58%', 'C2': '51.42%'}

{'F28': 'Z18', 'F13': 'Z14', 'F23': 'Time', 'F15': 'Z1', 'F2': 'Z19', 'F10': 'Z10', 'F6': 'Z4', 'F17': 'Z3', 'F26': 'Z12', 'F21': 'Z16', 'F7': 'Z7', 'F19': 'Z11', 'F8': 'Z9', 'F22': 'Z6', 'F11': 'Z23', 'F16': 'Z5', 'F18': 'Z17', 'F14': 'Z21', 'F20': 'Z24', 'F29': 'Z8', 'F27': 'Amount', 'F1': 'Z20', 'F30': 'Z27', 'F25': 'Z25', 'F24': 'Z13', 'F9': 'Z2', 'F5': 'Z22', 'F12': 'Z28', 'F3': 'Z26', 'F4': 'Z15'}

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

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

Fraud

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

GradientBoostingClassifier

C2

Printer Sales

Although the case under consideration has variables with a significant negative impact, it also has many measurable variables that are positive, so there is a good chance that C2 is correct since it has a 91.95% certainty. F3, F17, and F11 are the most important input variables, thanks to which the model successfully assigns the selected label, C2. F19 and F2 have almost identical positive impacts, while F8 has negative effects, shifting the output decision in favour of a different label. However, the cjoint positive contributions of F19, F3, F11, and F2 was higher than that of F17, F25, F7, and F8, increasing the likelihood of the C2 class. Unfortunately, the values of the variables F4, F6, F16, and F12 are likely ignored since their attributions are much closer to zero.

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

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

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

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

More

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

LogisticRegression

C4

Air Quality Prediction

The classification output observations that follow are based on the information supplied about this specific case. The class label in this case is forecasted to be C4 out of the four possible labels, with a probability of around 83.08 percent. With a probability of 16.87 percent, C1 is the next most likely label. The third possible label, C3, has a 0.05 percent chance of being correct. The algorithm, on the other hand, confirms that C2 is unlikely to be the correct label. According to the attribution analysis, F2, F3, F5, F6, F1, F4 is the ranking of the input features based on how powerful their effect on the algorithm is. Furthermore, among the input variables, F2 and F6 exhibit negative attributions, causing the decision to be shifted away from label C4. Finally, F3, F5, F1, and F4 are the positive variables that sway the judgement in favour of C4.

{'C4': '83.08%', 'C1': '16.87%', 'C3': '0.00%', 'C2': '0.05%'}

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

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

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

Preparing meals

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

DecisionTreeClassifier

C2

Insurance Churn

The model predicted class C2 with a very high confidence level of 93.27% and looking at the predicted probabilities across the label, there is only a 6.73% chance that C1 is the true label. There are two features that have a very strong positive effect on the prediction of class C2 and these are F9 and F11. The following features have moderate impact and are listed in descending order of impact: F12 and F4 have a negative impact, while F8 and F3 have a positive impact on the prediction of C2. In addition, both F5 and F13 had a negative effect on the model, further decreasing the odds of C2 being the true label for the given case. Finally, in terms of model decisions for this case, the features with the least contributions are F2, F7, F14, and F15.

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

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

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

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

Leave

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

RandomForestClassifier

C3

Flight Price-Range Classification

Between the three possible classes, there is a 100% certainty that the correct label for this case is C3. The features with a very high impact on the prediction made here are F7, F1, and F2, which are also shown to have a very strong positive contribution to the C3 prediction. Other features that shift the prediction in favour of C3 are F9, F8, F12, F10, and F6. On the other hand, F11, F3, and F5 negatively swing the model towards predicting a different label. Compared to F7, F1, and F2, all the negative features have a low to moderate influence on the prediction made here. Finally, F4 has the lowest positive contribution that also further increases the likelihood of the output label, C3.

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

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

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

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

Low

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

SVC

C1

Advertisement Prediction

For the given case or instance, the model assigns the label C1, with the prediction confidence equal to 56.56%. The variables F2, F1, F6, and F7 all contribute a lot to the classification decision above. While F2 and F6 are impacting positively, F1 and F7 are decreasing the likelihood of the assigned label. For the remaining features, both F4 and F5 shift the classification towards C1, whereas F3 has a marginal influence on the model, shifting the final verdict away in favour of the alternative label.

{'C1': '56.56%', 'C2': '43.44%'}

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

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

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

Skip

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

RandomForestClassifier

C2

Paris House Classification

The prediction made for this case by the model is that C2 is most likely the true label, with a confidence level of 72.03% higher than the 27.97% of the C1 label. According to the input features attribution analysis conducted, the features with the most influence on the decision are F15, F11, F8, and F7, all of which increase the probability that C2 is indeed the true label. The top negatively contributing features, increasing the probability that perhaps the true label could be C1, on the other hand, are F1, F5, F16, and F4. Conversely, F2, F17, F6, and F13 also have positive contributions, further pushing the decision towards labelling the case as C2. Overall, the fairly high confidence in the classification decision here can be attributed to the fact that positive features have a much higher influence on the decision than their negative counterparts.

{'C1': '27.97%', 'C2': '72.03%'}

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

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

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

Luxury

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

LogisticRegression

C1

Annual Income Earnings

Tasked with labelling cases, the classification model labels the case under consideration as C1 since the probability of C2 is only 20.22%. The predicted probability of the less probable class, C2, reflects the fact that the model is a bit doubtful about the output label. Responsible for this doubt are the negative features F4, F6, F5, and F10 since they support labelling the given case as C2 over C1. On the contrary, F2, F1, F14, F9, F12, F7, F3, and F8 are among the positively contributing features, responsible for the moderately high confidence in the classification output decision.

{'C2': '20.22%', 'C1': '79.78%'}

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

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

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

Above 50K

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

SVM_poly

C2

Mobile Price-Range Classification

The classification algorithm determines that neither C1 nor C3 nor C4 is a suitable label for the present context. C2 is quite guaranteed to be the correct label. The aforementioned conclusion has a higher degree of confidence due to the positive contributions of F18, F12, and F7. Aside from the above mentioned positive variables, F4, F5, F6, and F14 are also positive. However, their influences are moderate compared to F18, F12, and F7 . The remaining positive variables, F10, F20, F1, and F8, are among the algorithm's least influential input variables. Other attributes, such as F17, F19, F2, and F11, merely serve to reduce the likelihood of C2 being the proper label in the current context. Given the algorithm's high confidence in this classification, one may conclude that the negative variables had minimal impact on the algorithm's label selection here.

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

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

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

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

r4

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

SVM_linear

C1

Employee Promotion Prediction

The model gave the output label as C1 with a very high probability of 99.69%, leaving only 0.31% chance that C2 could be the right one. According to the contributions or attributions analysis done to understand the properties of various traits, F8 is by far the most influential trait. F7 had a positive impact on model predictions, as did F9. This is in contrast to F2 and F11, which have a negative impact on the model, pushing the classification verdict towards C2. Several input features are shown to have a limited impact on the output label produced by the model and they are: F1, F4, F5, F6, and F3. Overall, only the features F10, F2, F11, F1, F4, and F6 showed negative attributions, reducing the likelihood of the C1 label being assigned by the model but their joint impact was not enough to predispose the model toward a different classification decision.

{'C2': '0.31%', 'C1': '99.69%'}

{'F8': 'avg_training_score', 'F10': 'department', 'F7': 'KPIs_met >80%', 'F2': 'recruitment_channel', 'F11': 'age', 'F9': 'no_of_trainings', 'F1': 'previous_year_rating', 'F4': 'education', 'F5': 'region', 'F6': 'length_of_service', 'F3': 'gender'}

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

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

Promote

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

RandomForestClassifier

C2

Credit Risk Classification

Between the two classes, the given case is assigned the label C2 given that it has the highest predicted probability of about 93.0% since the probability of having C1 as the label is only 7.0%. Analysing the prediction made for the case under consideration, F6, F5, F7, and F3 are the features mainly pushing the prediction higher away from C2, while F2, F4, F1, and F8 improve the odds of the prediction being equal to C2. All things considered, the most relevant feature is F2, by contrast F10 and F9 are the ranked as the least relevant for the label assignment above.

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

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

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

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

Low

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

SVM_linear

C1

Wine Quality Prediction

The classifier says that C1 has a 67.54 percent chance of being the correct label for the given example or case; consequently the label C2 has a 33.46 percent chance of being the chosen class. The variables F8, F7, F6, and F9 have the most impact on the prediction judgement here. On the other hand, F10, F2, and F5 are seen as less relevant variables when determining the proper class. The variables F9, F4, F2, and F5 lower the probability of the assigned label C1 since they are negative variables favouring the C2 prediction decision. However, the other features' collective or joint attribution is strong enough to favour C1. In summary, F7, F8, and F6 are the most positive variables.

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

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

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

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

high quality

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

SVC

C1

Australian Credit Approval

The classification algorithm labels the presented data as C1 with the degree of confidence equal to 81.43 percent, although there is an 18.57 percent possibility that C2 is the correct label. The positive effects and contributions of input variables F12, F3, and F11 are mostly used to assign C1 to a specific scenario. Furthermore, the bulk of the remaining input variables contribute positively, making label C1 even more predictable. The only variables with negative contributions are F5, F1, F7, and F4, which move the choice to C2 rather than C1. Comparing the negative attributions to the positive attributions illustrates why the algorithm is certain that C1 is the correct label here.

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

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

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

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

Class 2

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

RandomForestClassifier

C2

E-Commerce Shipping

The probability that the label is C2 is 51.62% and the probability that C1 is the correct label is 48.38%. For this case or example, the uncertainty of the model is mainly due to the direction of influence of the variables F4, F5, and F6. Reducing the chance that C2 is the correct label are variables F4, F6, F10, and F8. While F4, F6, and F10 have a strong negative impact, F8 has the least negative contribution. Per the attribution analysis, increasing the prediction probability of C2 are the variables F5, F7, and F9 which are supported by F2, F3, and F1 all with moderate positive influences on the classification decision made by the model.

{'C2': '51.62%', 'C1': '48.38%'}

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

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

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

On-time

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

SVM_linear

C2

Wine Quality Prediction

The classifier assigns the label C2 since the probability associated with C2 is greater than that of C1. For the case under consideration, F4, F6, F9, and F2 are the sets of features significantly influencing the decision made by the classifier. However, features such as F7, F3, and F11 have limited to no impact on the classifier's output decision. F6, F4, and F2 are the features that are positively shifting the verdict toward predicting C2, not C1. In contrast, F9, F1, F3, and F11 have negative attributions, implying that they decrease the likelihood of C2 in favour of C1.

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

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

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

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

high quality

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

KNeighborsClassifier

C1

Cab Surge Pricing System

The probability that C3 is the label for the given case is zero and judging by the predicted probability associated with the remaining classes, the classifier is fairly certain that the correct label is C1 given its likelihood of 75.0%. The features are ranked in order of their respective impacts, from most important to least relevant: F8, F11, F4, F12, F2, F7, F1, F3, F6, F10, F9, and F5. Examining the contributions of the input features revealed that the ratio of negative features is smaller than the number of positive features. The negative features, F1, F7, F12, F4, and F3, decrease the classifier's response towards the generated class but the F8 value has the strongest positive contribution, increasing the response of the classifier to support the C1 assignment. Lastly, the least ranked features, F6, F10, F9, and F5, have a weak positive effect on the above prediction outcome, further increasing the odds in favour of label C1.

{'C2': '25.00%', 'C1': '75.00%', 'C3': '0.00%'}

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

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

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

C2

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

BernoulliNB

C2

Personal Loan Modelling

The most likely label for the given example based on the values of the variables is C2, according to the prediction probability of each class label. It can be concluded that the classifier is quite certain that C2 is the correct label because the probability of C1 is small. According to the attributions of the input variables, the most relevant features with a strong impact on the classifier's decision here are F9, F8, and F6, while on the contrary, the least relevant variables are F7 and F1. F9, F5, and F6 are positive variables that boost the classifier's response in favour of C2. The primary negative variables are F8 and F3, however they have a little impact on the above classification when compared to F9. Because the majority of the influential features have a positive impact, the confidence level of the classifier used to make the classification decision is high.

{'C2': '99.99%', 'C1': '0.01%'}

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

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

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

Reject

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

BernoulliNB

C2

Hotel Satisfaction

The given case is labelled as C2 since it has a prediction probability of 98.33% which implies that C1 is the least probable label. The higher confidence in the assigned label is mainly due to the contributions of input features F8, F5, and F15. In contrast, F7, F9, and F11 are the least ranked features. Based on feature attribution analysis, the top features F8, F5, and F15 have a strong positive influence, increasing the response of the classifier to assigning the label C2. Furthermore, pushing the decision further towards C2 are the other positive features such as F12, F10, F13, and F6. Supporting the prediction of the least probable class are the features F2, F1, F4, F14, and F11. When you compare the joint influence of the negative feature to that of the positive feature, it is evident why the classifier is very certain that C2 is the most probable label.

{'C1': '1.67%', 'C2': '98.33%'}

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

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

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

satisfied

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

KNeighborsClassifier

C1

Cab Surge Pricing System

The correct label, according to the classifier, is neither C2 nor C3, but C1, with a prediction likelihood of about 75.0%. By analysing the attributions of the input features, they can be ranked according to the level of impact, from the most important feature to the least relevant, as follows: F3, F7, F8, F6, F10, F4, F9, F2, F12, F11, F1, and F5. Among the twelve features considered by the classifier for the prediction verdict, seven have a positive influence on the classifier. F8, F6, F9, F4, and F2 are the five negative features that swing the assessment decision towards other classes. The value of F3 has a strong positive contribution to increasing classifier's response, favouring the assigning of C1. The last four features, F12, F11, F1, and F5, have a weak positive effect on the classifier's prediction for this case.

{'C2': '25.00%', 'C1': '75.00%', 'C3': '0.00%'}

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

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

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

C2

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

KNeighborsClassifier

C2

Basketball Players Career Length Prediction

It is important to note that the classifier's labelling decision is based solely on the information supplied. The classification verdict is as follows: C2 is the most probable label with respect to the case under consideration, since the prediction likelihood of the other label, C1, is only 12.50%. The most important variables contributing to the abovementioned classification are F5, F3, and F1, whereas remaining variables such as F2, F9, F18, F6, and F17 have a modest effect on the classifier's labelling decision for the given case. All the top features positively support the selection of C2 as the correct label and the negative variables increasing the chances of C1 are F11, F13, and F4. Given that these are the variables reducing the classifier's response towards generating label C2, it is not surprising that the classifier is very confident that C2 is likely the true label. In addition, the joint negative attribution of F11, F13, and F4 is very small when compared with the positive attributions of F5, F1, F2, and F3.

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

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

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

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

More than 5

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

DecisionTreeClassifier

C1

Concrete Strength Classification

According to the classification algorithm or model, C1 is the most likely class, with a very high confidence level, and C2 has a very low likelihood of being the right label. All of the inputs are proven to contribute to the categorization described above and the following is a ordering of the features from least essential to most significant based on their degree of influence: F1, F2, F5, F6, F3, F8, F7, and F4. It is clear from the attributions of the input attributes that the algorithm is quite certain that C2 is not the proper label for the given case since each attribute contributes positively, resulting in a significant push towards C1.

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

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

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

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

Weak

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

RandomForestClassifier

C1

Mobile Price-Range Classification

The model indicates that C3 and C4 have zero prediction probabilities, while that of C2 is 3.85%, meaning the most probable label for the given case is C1 and the confidence level is approximately equal to 96.15% certainty. The major features driving the above classification are F3, F20, and F14, while the least relevant features are F6, F1, F4, F18, and F17. The intermediate features have varying degrees of influence, from moderate to low, and these include F9, F10, and F8. Among the top influential features, only F9 has a negative contribution, driving the prediction slightly towards one of the other possible classes. Furthermore, the top two positive features, F20 and F3, have a stronger influence than all the negative features combined. It is, therefore, not surprising that the model is confident about the classification verdict here.

{'C3': '0.00%', 'C4': '0.00%', 'C2': '3.85%', 'C1': '96.15%'}

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

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

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

r4

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

SVM_poly

C1

Mobile Price-Range Classification

The predicted output label from the model is C1 with almost 100% certainty, indicating it is very certain it is correct and this is mainly because the likelihoods across the other labels C4, C2, and C3 are 0.47%, 0.05%, and 0.04%, respectively. Among the top features F19, F11, and F6, the features F6 and F11 positively influence the classification decision above in the direction of C1, whereas F19 influences in the opposite direction in favour of an alternative label. With a similar direction of influence as F19, the features F7, F17, F5, and F8 negatively impact the prediction of C1, whereas F2 positively impacts it. Features F3, F17, F18, and F1 also have a smaller influence on the prediction output for the given case and finally, the features F10, F4, and F13, have very little contributions to the classification made by the model for the case under consideration.

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

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

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

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

r1

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

SVC

C1

Paris House Classification

The prediction probabilities associated with the classes C1 and C2 are 99.56% and 0.44%, respectively. Therefore, we can conclude that the most probable label for the given data is C1. The classification model's decision here is largely based on the impacts of the F9, F6, and F10, whereas the F8, F7, and F2 have very little to say about the decision here. In terms of the direction of influence of the features, F9, F16, F12, F13, and F5 are the top positive features contributing to the prediction outcome of C1. Conversely, the marginal doubt in the classification decision (represented by the probability of C2) is largely due to the negative contributions of F6, F4, F14, and F3. To sum up, the very high certainty in the classification output decision could be explained by considering the fact that the joint influence of the negative features is smaller than that of the positive features.

{'C1': '99.56%', 'C2': '0.44%'}

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

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

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

Basic

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

GradientBoostingClassifier

C2

Basketball Players Career Length Prediction

The model identifies the case as C2 since, the true label has just 33.63 percent chance of being C1 when the prediction probability is calculated. The in-depth analysis found that the bulk of the attributes had negative impacts, driving the prediction away from C2 and toward C1. F12, F17, F11, F18, and F9 are among the features that contribute negatively. Furthermore, these features' values are ranked higher than any of the positive features, which are F13, F2, F8, and F15. Finally, it can be concluded that the values of F5, F19, and F16 are less important in predicting the outcome of the case under review, hence they are ranked the least.

{'C1': '33.63%', 'C2': '66.37%'}

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

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

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

Less than 5

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

KNeighborsClassifier

C2

Cab Surge Pricing System

0.0% is the predicted probability that C1 is the true label for the test example under consideration according to the classifier. Judging based on the predicted probabilities associated with the other remaining labels, the classifier is 75.0% confident that C2 is the correct label. From the analysis, the features ranked according to the degree of impact from the most significant feature to the least relevant ones: F8, F3, F1, F5, F12, F9, F4, F10, F7, F11, F2, and F6. Examining the contributions or attributions of the features further revealed that the ratio of positive features to negative features is seven to five. The negative features swinging the prediction decision towards the other classes are F1, F5, F4, F9, and F10 since their contribution decrease the probability that C2 is the true label for the given case. The value of F8 has the strongest positive contribution increasing the classifier's response in support of assigning C2 but the last four features, F7, F11, F2, and F6, have a weak positive influence on the labelling decision or conclusion with respect to the given case.

{'C3': '25.00%', 'C2': '75.00%', 'C1': '0.00%'}

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

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

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

C2

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

SGDClassifier

C1

House Price Classification

The classifier is very certain that C2 is not the true label since the predicted probability of C1 is given as 100.0%. Analysing the attributions of the features indicates that the most relevant features are F2, F4, F12, and F11 while F9, F6, and F8 are the least relevant features. The values of F7, F1, F3, F5, F10, and F13 have a moderate influence on the classification decision made here. Considering that the classifier is 100.0% certain that C1 is the true label, we can conclude that the collective negative attribution of F12, F1, and F13 is clearly outweighed by the positive attributions of features such as F2, F4, F7, and F11.

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

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

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

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

Low

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

GradientBoostingClassifier

C1

Health Care Services Satisfaction Prediction

Based on the information provided to the classifier, the true label for the given case is likely C1, with a confidence level of 76.26%. Each input variable has a different degree of influence on the classifier's final labelling decision with respect to the case under consideration. Whilst F7, F12, and F5 have lower contributions to the classifier's decision, F8, F9, and F11 are identified as the major contributors resulting in the assignment and classification probabilities across the two classes. There is a 23.74% chance that perhaps C2 is the true label and the features responsible for this are the negative features, F11, F6, F2, F14, F10, F7, and F12. Driving the classifier's decision in favour of C1 are the positive features such as F8, F9, F3, F16, F13, F1, and F15.

{'C2': '23.74%', 'C1': '76.26%'}

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

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

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

Satisfied

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

SVC

C2

Health Care Services Satisfaction Prediction

The classification model employed made its label selection decision based on the information provided about the case under consideration. With a moderately low degree of confidence, it classifies the case under consideration as C2. Specifically, per the model, the probability of labelling the case as C1 is equal to 48.66%, hence not as likely as C2. The decision made here can be attributed to the influence of features such as F4, F11, F13, F5, and F2. However, F3, F6, F16, F1, and F10 are the least relevant features with respect to the classification made. The confidence level of the model is marginally above average and this can be attributed to the negative contributions of F9, F4, F15, F14, F3, F1, and F10. The negative features shift the prediction decision in the direction of C1, however, the positive contributions of other features such as F11, F13, F5, and F2 improve the odds of the C2 label.

{'C1': '48.66%', 'C2': '51.34%'}

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

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

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

Satisfied

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

SVC

C2

Flight Price-Range Classification

The prediction results are as follows: the probability that C2 is the correct label is 97.12%, the probability that C1 is the correct label is 2.55%, and the probability that C3 is the correct label is 0.33%. Judging based on the prediction probabilities across the classes, C2 is the most probable label. The very high confidence in the assigned label can be attributed to the very strong positive influence and contributions of the variables F8, F5, F9, F6, and F2. The other positive variables are F4, F7, and F11. The positive variables increase the probability that C2 is the correct label for the given case. Decreasing the probability of C2 are the negative variables F12, F1, F3, and F10. Considering that the combined effect of the negative factors is quite minimal in comparison to the top positive variables, it is not surprising that the model is very sure that neither C1 nor C2 is the best label for the given case.

{'C2': '97.12%', 'C1': '2.55%', 'C3': '0.33%'}

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

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

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

Low

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

DecisionTreeClassifier

C2

Insurance Churn

C2 is the model's predicted output for this given case, with an accuracy of 87.13% meaning the likelihood of C1 is only 12.87%. F9, F12, F10, F7, and F16 have the most effect on the output prediction choice in this case, whereas on the other hand, F14, F6, F2, and F11 are not that important to the decision made here. F9, F16, and F7 are the top negative features when you consider direction of their respective impacts, decreasing the model's reaction to labelling the given scenario as C2 and also F13, F5, F6, F2, and F11 are the other features that contribute negatively. In a nutshell, F12, F10, F1, F4, and F8 are primarily positive improving the odds of C2 with respect to this classification conclusion.

{'C2': '87.13%', 'C1': '12.87%'}

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

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

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

Stay

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

SVM_linear

C2

Employee Promotion Prediction

The classification model or algorithm classifies the provided data or case as C2 with a predicted likelihood of 94.16%, meaning that the chance of C1 being the true label is only 5.84%. The most relevant features driving the classification above are F2, F5, F3, F4, and F10, however, arranging the input features in-order of their contributions revealed that the least influential features are F7, F8, F1, and F6 since their values receive little consideration or emphasis from the algorithm. In relation to the directions of influence of input features, only F4 and F1 are shown to have negative contributions, which tends to drive the labelling judgement towards C1 instead of C2. Considering that the combined effect of all the negative features is lower than that of the positive features such as F2, F5, F3, F10, F11, and F9, it is valid to say that C2 is the most probable label.

{'C1': '5.84%', 'C2': '94.16%'}

{'F2': 'department', 'F5': 'avg_training_score', 'F3': 'KPIs_met >80%', 'F4': 'recruitment_channel', 'F10': 'region', 'F11': 'education', 'F9': 'length_of_service', 'F8': 'age', 'F1': 'no_of_trainings', 'F6': 'gender', 'F7': 'previous_year_rating'}

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

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

Promote

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

GradientBoostingClassifier

C1

Basketball Players Career Length Prediction

The final classification made was C1, but with a likelihood of only 55.19%, the model is uncertain about this prediction. By far, feature F10 had the most impact and following F10 are F1, F13, and F19 have been identified as having the comparable influence on classification. The combination of F10, F1, F13, F19, and F15 features has shifted the classification decision from C1 to C2. While F18, F6, and F12 are all features with a moderate impact on the classification, F18 is the only one of that set that has had a positive impact on the C1 classification and the remaining positives are F14, F17, and F9. Lastly, the features F16, F2, F3, F5, and F11 had very marginal negative contributions to the classification verdict.

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

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

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

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

Less than 5

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

BernoulliNB

C2

Personal Loan Modelling

Based on the prediction probabilities, C2 is the most likely label for the given case considering the values of the input variables and because the likelihood of C1 is very marginal, so the classifier is very confident that C2 is the right label. An analysis of the contributions of the variables has shown that F3 is the most relevant, with the strongest influence on the classifier's decision, however, to arrive at the classification above, the classifier probably ignores the values of the least ranked variables, F4 and F9. The level of confidence of the classifier with respect to the above classification decision is higher, primarily because most of the influential variables have a positive impact. F3, F5, and F1 are the top positive variables that increase the likelihood of C2. Having a different direction of influence, F8, F9, F4, and F2 are the negative factors, but compared to F3, their impact on the prediction decision above is low.

{'C2': '99.99%', 'C1': '0.01%'}

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

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

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

Reject

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

KNeighborsClassifier

C2

Real Estate Investment

Based on the information available about the case under consideration, the classification model is very uncertain about the appropriate labels for the case. According to the model, there is an almost equal distribution in terms of the probability that any one of C2 and C1 is an appropriate label. This indicates that any of the possible labels could be the true one, but for simiplicity, the model selects the class as C2. The above judgement is mainly due to the influence of the following factors or variables: F3, F19, F13, and F10 while the least relevant variables are F17, F1, and F7. Positive variables like F10, F13, F5, and F20 increase the model's response in favour of the assigned label. Nevertheless, negative variables such as F3, F4, F11, and F19 reduce the possibility that C2 is an appropriate label because their values support the selection of C1. Uncertainty about the classification here can be due to the fact that the most important negative properties, F3 and F19, have very high impacts, which moves the model's judgement away from C2 towards C1.

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

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

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

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

Ignore

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

LogisticRegression

C2

Tic-Tac-Toe Strategy

There is about an 81.01% chance that C2 is the probable label, hence the predicted probability for the C1 class is only 18.99%. The algorithm or classifier arrived at the prediction verdict above mainly based on the influence of features such as F3, F9, F1, and F2. For the algorithm, the least relevant feature is F5, which is shown to have a very small contribution in relation to the label choice here. When the directions of influence of the input features were investigated, it was discovered that F3, F8, F1, and F2 have positive attributions, pushing the algorithm higher towards the C2 label. Negative features such as F9, F7, and F4 assist in dragging or pushing the classification decision lower towards C2, where it was originally classified and this is mainly because their contributions to the prediction favour choosing or labelling the case as C1.

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

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

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

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

player B win

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

BernoulliNB

C2

Hotel Satisfaction

According to the classification algorithm, there is 77.69% chance that the given case is part of the C2 population. The features with the largest impact driving the algorithm to arrive at the above decision are F13, F8, and F2 which are followed in the decreasing order of influence by F11, F7, F5, F10, F4, F15, F14, F1, F12, F9, F3, and F6. Inspecting the direction of influence of the input features showed that, F13, F4, F15, F6, and F8 have negative influence on the prediction, shifting the algorithm's verdict towards the C1 class and can be blamed for the doubt in the classification decision. However, strongly pushing the classification higher towards the C2 label are the positive features such as F2, F11, F7, F5, F10, and F14.

{'C2': '77.69%', 'C1': '22.31%'}

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

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

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

dissatisfied

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

DecisionTreeClassifier

C1

Insurance Churn

The likelihood of the true label for the given test case being equal to the model's output prediction, C1, is 85.71% and since it's not 100%, there is a small chance of about 14.29% that the model could be wrong. Among the features employed for this classification, F12, F2, F8, F13, F9, and F16 are the top features influencing the model's prediction decision. The features with the strongest positive influence are F12 and F2 and in fact, these are shown to be the two main driving forces controlling the model's decision regarding the given case. Besides, some otf the other positive features include F13, F9, F15, F4, and F16. However, the atrribution of F8, F10, F7, F5, and F11 indicates the true label could perhaps be C2. While the different input features have some sort of contribution to the prediction made for this test case, the features F6, F3, and F1 have the least impact on the final decision here.

{'C2': '14.29%', 'C1': '85.71%'}

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

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

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

Leave

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

SVC

C2

Vehicle Insurance Claims

To begin with, the classification choice is entirely dependent on the information or data provided to the prediction model. According to the model, C2 has a 61.61 percent probability of being the true label, whereas C1 has a 38.39 percent chance of being the true label. Because the estimated probability of C2 is greater than that of C1, it is reasonable to assume that C2 is the most probable true label. The key variable responsible for this classification is F17, with a very significant positive effect on the model's conclusion, pushing it higher towards C2. F3, F21, F8, F1, F25, F16, F12, and F28 are the next set of relevant variables. F3, F8, F1, F16, F15, F33, and F12 have negative contributions that are responsible for the decrease in the chance that C2 is the actual label since they prefer to assign the C1 label instead. This means that the contributions of F21, F25, F9, F32, and F28, together with F17, can explain why the model is rather confident that C2 is the correct label.

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

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

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

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

Fraud

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

LGBMClassifier

C2

Employee Promotion Prediction

With a prediction likelihood of 62.34%, the model trained to generate predictions based on input variables identifies the presented example as C2. The model's label assignment choice for the given case is heavily impacted by the values of input variables such as F9, F11, and F7. The least important variables, on the other hand, are F1, F4, and F5. Furthermore, the impact of F3, F6, and F8 is regarded as moderate. F7 and F3 are the variables identified to have negative contributions to the classification when you take into consideration their respective direction of impact. All of the remaining variables have a positive influence, contributing to the classification of the presented case as C2. As a result, it is unexpected that the model's confidence is just 62.34% which suggest that the negative attributes may have a larger say in the appropriate label for the case under review.

{'C1': '37.66%', 'C2': '62.34%'}

{'F11': 'department', 'F9': 'avg_training_score', 'F7': 'recruitment_channel', 'F3': 'KPIs_met >80%', 'F6': 'no_of_trainings', 'F8': 'length_of_service', 'F2': 'age', 'F10': 'region', 'F1': 'education', 'F4': 'previous_year_rating', 'F5': 'gender'}

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

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

Promote

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

SVC

C2

Food Ordering Customer Churn Prediction

For the case under consideration, the model outputs C2 with high confidence level since the associated predicted class label is 89.73% whilst that of C1 is just 10.27%. Just few features out of the entire input features are shown to have control over the prediction made here. The prediction verdict C2 is mainly based on the variables F28, F18, F7, and F45. Other variables with moderate attributions include F3, F2, F20, F29, F24, and F8. Each variable mentioned above is shown to have different direction of contribution or impact for instance while F28, F45, F24, and F29 positively support the model's output decision, F18, F7, F3, F2, F8, and F20 contributed to decreasing the likelihood or odds of C2 being the true label for the given test instance. The variables shown to have no influence or contribution on the classification decision above are mainly F19, F9, F37, and F36.

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

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

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

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

Return

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

RandomForestClassifier

C2

Health Care Services Satisfaction Prediction

The model trained to solve the classification task labels the given case as C2 with a moderately high degree of confidence level equal to 60.13%. However, it is important to note that the prediction likelihood of C1 is 39.87%. Investigation of the contributions of the features to the above label assignment indicates that the most relevant features considered by the model are F6, F7, F8, and F5. Increasing the prediction likelihood of label C2 are mainly the positive features F6, F8, and F5. These features are termed positive features since their direction of influence is in support of the assigned label C2. On the contrary, F7, F16, and F13 are the top negative features, accounting for the uncertainty in the final prediction verdict. In plain terms, these negative features support labelling the case as C1, contradicting the model's decision in this case.

{'C2': '60.13%', 'C1': '39.87%'}

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

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

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

Dissatisfied

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

DNN

C2

Concrete Strength Classification

The following assertions are based on the information provided to the classification model. The classification model's confidence in this case's prediction output is approximately 69.40% and this suggest that the chance of label C1 is about 30.60%. The prediction attribution analysis shows that F3 and F6 are the most important features, whereas F8 and F1 are the least influential. F5, F4, and F2 are recognised as the only negative features considering the direction of effect of the features since their contributions reduce the prediction likelihood of the specified label, C2. F3, F6, F7, F8, and F1, on the other hand, have a positive impact on the model in favour of labelling the provided situation as C2 rather than C1.

{'C2': '69.40%', 'C1': '30.60%'}

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

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

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

Weak

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

GradientBoostingClassifier

C2

Broadband Sevice Signup

Because the chance that the label is the alternative class C1 is only 1.94 percent, the model anticipates that C2 will be the correct label in this situation. Specifically, it can be concluded that the model has a high level of confidence in the label C2. The feature attribution analysis conducted suggests that the two most relevant features considered when choosing the C2 are F12 and F33. F10, F3, F23, F31, and F27 were some of the other factors that positively helped with this prediction. F28, F30, F42, and F35, on the other hand, are the features with a negative influence on the above prediction judgement. In comparison to the F14, F27, F3, and F33, the foregoing features have little impact on the model and this might explain why the model is so certain that the correct label is C2. However, it is crucial to note that not all features are considered by the model during the label assignment with the irrelevant features such as F34, F11, F20, and F24 having extremely low attributions which happens to be almost zero.

{'C2': '98.06%', 'C1': '1.94%'}

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

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

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

No

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

RandomForestClassifier

C2

Student Job Placement

The model predicted that the example should be classified as C2 with a 76.06% likelihood but the model also identified that there was a 23.94% chance that the right label could actually be C1. The positive influence of features F12, F11, F1, and F9 on the model supports the class assignment of C2. Both F5 and F6 are features with a small positive impact on the classification decision for the given case. F10 and F2, in contrast, has a small negative impact on the output verdict that drives the decision away in favour of the other label. The features F7 and F4 have only a very small impact on the final classification decision. Finally, F3 is shown to have zero impact on the model in this case, hence it is not relevant to the prediction of class C2.

{'C2': '76.06%', 'C1': '23.94%'}

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

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

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

Not Placed

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

RandomForestClassifier

C1

Used Cars Price-Range Prediction

The classification model labels the given case as C1 at a very high confidence level since the probability that C2 is the correct label according to the model is only 3.50%. The assignment decision above is mainly based on the values of the features F8, F2, F1, and F7. On the other hand, the values of F6 and F3 are shown to have a very weak influence on the model's decision. The analysis revealed that only four of the input features support the decision by the model, while the remaining ones contradict the assigned label. The four positive features are F1, F7, F10, and F5.

{'C2': '3.50%', 'C1': '96.50%'}

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

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

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

High

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

SVM_linear

C1

Employee Promotion Prediction

The model generated the label, C1, with a very high likelihood of 99.69%, hence the probability that C2 is the right label is only 0.31%. Based on the analysis performed to understand the attributions of the different features, F9 was by far the most impactful positive feature whereas, the most negative feature is identified as F7. F2 also had a positive influence on the model's prediction, as did F8, F6, and F1. This is in contrast to F3 and F5, which had a negative influence on the prediction. Many of the features under consideration had only smaller impact on the outcome of the model and these are F4, F10, F1, F11, and F6. Considering the attributions of the input features, only F7, F3, F5, F4, F10, and F11 are shown to have negative attributions, decreasing the likelihood of the predicted label, however, the collective influence of the negative features is not enough to swing the model towards a different label.

{'C2': '0.31%', 'C1': '99.69%'}

{'F9': 'avg_training_score', 'F7': 'department', 'F2': 'KPIs_met >80%', 'F3': 'recruitment_channel', 'F5': 'age', 'F8': 'no_of_trainings', 'F4': 'previous_year_rating', 'F10': 'education', 'F1': 'region', 'F11': 'length_of_service', 'F6': 'gender'}

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

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

Promote

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

KNeighborsClassifier

C1

Advertisement Prediction

The item is labelled as C1 with a high degree of confidence since the predicted probability associated with the other class is 0.0%. Looking at the contributions of the features, only F3 and F5, are shown to drive the model towards predicting C2. However, these features are ranked as the least relevant, implying that their values have a very low impact on the model's decision. All the positive features, F4, F7, F2, F1, and F6, are ranked higher than the negative ones, with higher impacts on the model, significantly supporting the assigned label which could explain the high confidence level.

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

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

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

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

Skip

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

SVC_linear

C2

Personal Loan Modelling

With the prediction probability distribution across the labels, C1 and C2, respectively, equal to 0.30% and 99.70%, the model labels this instance as C2. The most important features are F4, F6, and F7. The variables, F3, F5, F1, and F2, have values, increasing the chances of C1 being the label for this case. Increasing the odds of C2 being the correct label are the values of the remaining variables. The strong positive variables are F4, F6, and F7 coupled with the moderate positive influence of F9 and F8 pushes the prediction in favour of C2 hence the prediction confidence level achieved.

{'C1': '0.30%', 'C2': '99.70%'}

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

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

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

Accept

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

DecisionTreeClassifier

C1

Car Acceptability Valuation

C1 is given as the predicted label with very high confidence, and according to the classification algorithm, there is no chance that either of the remaining three labels, C4, C1, and C3, is the right label for this case since the predicted probability of C2 is 100.0%. Based on the attribution analysis and investigations, the ranking of the input features from the most important to the least important is: F4, F5, F6, F1, F2, and F3. From the attribution analysis, F4 is the only one that positively contribute and support the above classification decision, while the remaining features such as F5, F6, F2, and F1 have negative contributions, shifting the decision in a different direction. In conclusion, looking at the predicted confidence level, one can say that the very strong attribution or influence of F4 is enough to dwarf the contributions of the features F5, F6, F1, F2, and F3.

{'C2': '100.00%', 'C4': '0.00%', 'C3': '0.0%', 'C1': '0.0%'}

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

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

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

Unacceptable

{'C2': 'Other B', 'C4': 'Acceptable', 'C3': 'Other A', 'C1': 'Unacceptable'}

KNeighborsClassifier

C2

German Credit Evaluation

In the present case, there is only a 12.50% chance that C1 is the correct label, which means there is an 87.50% chance that C2 is the true label. Therefore, the most probable class assigned by the model is C2. The above decision is mainly based on the influence of the following variables: F1, F2, and F6. Of these main variables, only F2 had a very strong positive impact on the model, increasing the prediction probability of the assigned label. The most important variables that lower the likelihood of C2 being the correct label are F6 and F1. The remaining two variables moving the decision away from C2 are F9 and F3. F5 and F4 are the least important variables, with a marginal impact on the model and this positive impact on the model is moderately low.

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

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

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

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

Good Credit

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

KNeighborsClassifier

C2

Tic-Tac-Toe Strategy

There is an evenly split chance that the prediction could be either of the two labels, C2 and C1. Based on the predicted probabilities, we can conclude that the model is uncertain about which label is the correct one. The abovementioned prediction decision is chiefly attributed to the influence of the following features: F8, F1, and F5, however, the least important or ranked ones are F6 and F4. The attributes F3, F7, F9, and F2 are shown to have moderate contributions.

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

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

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

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

player B lose

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

SVC

C2

Water Quality Classification

The label assigned to the given sample is C2 at a confidence level of 56.81%. This means that there is a 43.19% chance that the sample could be C1, representing an uncertain classification decision. The values of F1, F8, F6, F7, and F2 are the major contributing factors resulting in the classification decision here. On the other hand, the least important features are F4, F9, and F5, with a low level of influence. Considering the direction of influence of the features (that is, either supporting or contradicting the prediction above), only F6, F7, and F2 are shown to have positive attributions, increasing the likelihood of the assigned label. This implies that the values of the remaining features F3, F1, F8, F9, F4, and F5 have negative attributions, shifting the verdict in the opposite direction in favour of C1. In simple terms, the correct label should be C1 according to the negative features enumerated above.

{'C2': '56.81%', 'C1': '43.19%'}

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

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

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

Not Portable

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

SVC

C2

Australian Credit Approval

The classification algorithm classifies the given case as C2, since there is only an 18.57% chance that C1 is the correct label. The effects and contributions of positive input variables F1, F8, and F10 are the major drivers for the above classification. Besides, most of the remaining predictors such as F13, F9, F4, F7, and F3, are positive variables, decreasing the likelihood of the C1 label and making the label C2 more likely. The only variables with negative contributions are F14, F3, F6, and F2, which motivate generating the label C1 instead of C2. In summary, comparing negative attribution to positive attribution explains why the algorithm can determine that C2 is the right label for the given case.

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

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

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

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

Class 2

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

LogisticRegression

C1

Concrete Strength Classification

The odds are in favour of C1 being the correct label for the given case. This is because the probability of the other label, C2, is only 1.03%. Ranking the features in order of relevance to the classification decision above, F1, F4, F3, F6, F5, F2, F8, and F7. Among the set of features used for this prediction, F4, F5, and F2 are the only ones shown to decrease the likelihood of the C1 decision. The positive features increasing the chances of C1 being the correct label are F1, F3, F6, F8, and F7. The joint attribution of the positive features is stronger than that of the negative ones, which explains the confidence level associated with class C1.

{'C2': '1.03%', 'C1': '98.97%'}

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

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

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

Strong

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

GradientBoostingClassifier

C2

Paris House Classification

According to the prediction algorithm or model, there is almost 100% confidence that C2 is the label for the case under consideration. This is because the probability of C1 being the correct label is only 0.70%. The classification decision above is mainly based on the values of the following features: F8, F17, and F10 since their respective attributions are higher than any of the remaining features. F17 has a negative contribution to the prediction made by the model for this case, while in contrast, F8 and F10 have positive contributions, that push the classification decision in favour of C2. Unlike all the features mentioned above, the values of F14, F16, F7, and F1 have a limited impact on the classification decision above.

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

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

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

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

Basic

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

MLPClassifier

C2

Ethereum Fraud Detection

The C1 has a predicted probability of just 3.10 percent, but the C2 has a predicted probability of 96.90 percent, which implies that C2 is the most likely class chosen by the classifier for the supplied data. Not all of the input features are directly relevant to labelling the provided data and, per the attributions analysis, only F9, F12, F27, F15, F35, F33, F32, F29, F3, F18, F13, F37, F1, F36, F17, F23, F8, F26, F2, and F21 are the relevant features. However, F16, F6, and F7 are examples of irrelevant features since their contributions are mostly ignored by the classifier when classifying the given case. According to the attribution assessment, F9 and F12 have a very substantial combined positive influence, enhancing the classifier's response towards C2 rather than C1. In contrast, the top negative features are F27, F35, and F15, which weaken the classifier's response in favour of C1. When the attributions of F9, F33, and F12 are compared to the attributions of the negative features indicated above, it is not unexpected that the classifier is highly certain that C2 is the most likely label in this case.

{'C1': '3.10%', 'C2': '96.90%'}

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

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

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

Fraud

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

RandomForestClassifier

C1

German Credit Evaluation

The classification algorithm labels this instance as C1, but its level of confidence is moderate considering the fact that there is about a 44.0% chance that C2 could be the appropriate label. The features, F5, F7, F9, and F2, negatively influence the prediction verdict away from C1 and favour assigning C2 as the correct label. Contradicting the influence of the negative feature are features such as F1, F4, and F6, with positive contributions, improving the odds in favour of the probable label, C1. To summarise, the top features with the most influence on the above label assignment are F1 and F5, but F2 and F3 are the least influential input features considered by the algorithm.

{'C1': '56.00%', 'C2': '44.00%'}

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

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

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

Good Credit

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

GradientBoostingClassifier

C1

Food Ordering Customer Churn Prediction

The case given is labelled as C1 with close to an 82.07% confidence level, implying that the likelihood of C2 being the correct label is only 17.93%. The classification above is mainly due to the contributions of different features such as F43, F8, F5, F11, F17, and F41. But, not all features are considered by the classifier to arrive at the decision made for the given case. These irrelevant features include F16, F35, F19, and F12. Among the influential features as shown, F43, F8, F5, F11, and F46 are the top positives that increase the probability of C1 being the true label. However, F17, F41, F45, F30, F6, F13, F2, and F26 are the top negative features, driving the prediction lower towards C1 in favour of C2. In closing, the most important features with regard to this classification output are F43 and F8, all with positive attributions, explaining the very high confidence level.

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

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

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

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

Go Away

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

RandomForestClassifier

C2

Company Bankruptcy Prediction

The output labelling decision is C2 with almost 100% certainty, which indicates that there is practically no chance that C1 is the right label choice for the case under consideration. F34, F82, F44, F81, and F11 are the features with the highest joint positive impact, influencing the model's decision to output C2 and the feature F65 also has a high impact, but unlike F34, F82, F44, F81, and F11, F65 attempts to shift the decision away from C2 in the direction of C1. Also, F71 and F30 have a moderate impact on the decision towards C2, although this is still higher than features F48, F7, F4, and F79, which have a moderate impact, favouring the prediction of class C1. Besides, F30, F38, F80, F12, F77, and F57 all have a positive influence on the final classification verdict further increasing the odds in favour of the C2 label. It is worthy to note that for this classification decision, a large number of features are shown to be irrelevant hence received negligible consideration from the model, and these include F5, F91, F36, F90, and F18.

{'C2': '99.00%', 'C1': '1.00%'}

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

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

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

No

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

LogisticRegression

C2

House Price Classification

The label assigned by the classifier in this instance is C2, which had a very high prediction likelihood of about 99.93%. According to this classifier, the probability of C1 being the correct class is only 0.07%. Analysis performed shows that the confidence level of the classifier here is due to mainly the values of the features F1, F10, F13, and F4. The least relevant features to this classification verdict are F11, F6, F2, and F12 since the magnitude of their respective attribution is smaller compared to the remaining features. Furthermore, only the features, F9, F5, and F6, have a negative influence, increasing the chances of predicting the alternative label C1. However, when compared to the joint influence of the positive features such as F1, F10, and F13, the influence of the negative features is smaller, hence explaining the high degree of confidence in the predicted C2 label.

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

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

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

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

High

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

BernoulliNB

C1

Credit Card Fraud Classification

The classifier is very certain that C2 is not the accurate label for the given data or example, but that C1 fits. F26, F23, F21, F30, F1, F17, and F5 are the input features that have the most influence on the choice or judgment. F6, F13, F9, F24, F8, F16, F4, F11, F27, and F15, on the other hand, are found to be irrelevant and have negligible inlfuence on the classifier. Amongst the top features, F26, F23, and F21 are the one shown to have negative contributions, greatly favouring C2, lowering C1's prediction probability. Despite the significant negative attributions of the top impactful attributes, the classifier is quite certain that C1 is the correct label, based on the prediction probabilities.

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

{'F26': 'Z4', 'F23': 'Z3', 'F21': 'Z23', 'F30': 'Z2', 'F1': 'Z10', 'F5': 'Z7', 'F17': 'Z12', 'F18': 'Z14', 'F10': 'Z24', 'F29': 'Z28', 'F3': 'Time', 'F2': 'Z19', 'F7': 'Z26', 'F28': 'Z16', 'F14': 'Z5', 'F22': 'Z22', 'F12': 'Amount', 'F19': 'Z9', 'F25': 'Z18', 'F20': 'Z15', 'F6': 'Z17', 'F13': 'Z1', 'F9': 'Z20', 'F24': 'Z21', 'F8': 'Z13', 'F16': 'Z11', 'F4': 'Z25', 'F11': 'Z8', 'F27': 'Z27', 'F15': 'Z6'}

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

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

Not Fraud

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

SGDClassifier

C1

House Price Classification

The classifier's anticipated label for this case is C1 which is a decision that it is highly confident about since the predicted likelihood is 100.0%. The most important variables are F1, F11, F13, and F12, whose values lead to the aforesaid classification conclusion. Under this classification instance, examination of the attributions of the features showed that F10, F3, and F2 are the least essential features. Because majority of the case's attributes positively validate the assigned label, it's not unexpected that the classifier picked the C1. F1, F11, F12, F5, F8, and F4 are all positive variables, while F13, F7, and F6 are three contradicting variables that moderately drive the labelling judgment towards C2.

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

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

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

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

Low

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

GaussianNB

C1

Tic-Tac-Toe Strategy

The model selects C1 as the correct label with a probability of 57.58%, while the other class, C2, has a slightly lower probability of 42.42%. The most relevant attribute is F9, followed by F1, F2, F5, F7, F3, F8, F6 and finally F4, which is the least relevant. The features F7, F8, and F9 have a positive influence, increasing the probability of the classification output, while F2 has a negative attribution, swinging the model to assign C2 instead. F5, F6, F1, and F3 are some of the other negative attributes. Finally, F4 has a very small positive control over the prediction in this test case but it further increases the confidence in the label chosen for the given case.

{'C1': '57.58%', 'C2': '42.42%'}

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

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

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

player B lose

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

SGDClassifier

C1

House Price Classification

The prediction verdict here is that the most probable class label is C1. Actually, the classification algorithm indicates that there is no possibility that the correct label is C2. Majorly contributing to the above classification are F2, F6, F11, and F10, all with positive influence. It is therefore not surprising that the algorithm is confident that C1 is the right label. The other positive features considered to arrive at the decision here are F12, F5, F1, F7, and F3. According to the attribution analysis, only F9, F13, and F4 have negative contributions, which tend to attempt to swing the final verdict in favour of C2. To sum up, the joint negative influence is not enough to outweigh the positive features, hence the C1 is assigned for the given case.

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

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

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

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

High

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

DecisionTreeClassifier

C1

Hotel Satisfaction

Due to the prediction probability distribution across the class labels, the labels assigned to this example is C1 with a high degree of confidence, close to 100 percent. The most significant features driving the classification above, according to the attributions of the input features, are F10, F1, F5, and F8. F9 and F15, on the other hand, are the least essential features to this prediction here. In addition, just four of the input features have a negative impact, skewing the classifier's judgement in favour of the C2 label. F9, F5, F4, and F15 are the opposing features. The contribution of the negative features, with the exception of F5, is quite modest when compared to the top positive features such as F1, F8, and F7.

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

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

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

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

satisfied

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

RandomForestClassifier

C2

Student Job Placement

The classification algorithm predicts that the data sample given should be classified as C2 with a probability of 76.06%, but it also finds that there is a 23.94% probability that the correct label will be C1. The positive influence of the F4, F1, F6, and F9 features on the algorithm supports the C2 class tasks. F7 and F11 are features with little positive influence on the classification decision for a particular case. F3 and F5, in contrast, has a small negative impact on the output decision that result in the reduction in the likelihood of C2 hence can be said to favour labelling the case as C1. F10 and F12 had only a minor positive impact on the final labelling decision and finally F2 was shown to have zero effect on the algorithm in this case.

{'C2': '76.06%', 'C1': '23.94%'}

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

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

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

Not Placed

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

KNeighborsClassifier

C2

Credit Risk Classification

According to the model, there is a higher chance that the case's label is C2. This prediction decision is based primarily on the attribution of the following features: F1, F2, F7, and F8. Aside from F8, all the other features listed above have a strong positive influence, increasing the probability of the predicted class C2. Similar to F8, the values of features F6, F3, and F10 suggest the other label, C1, could be the correct label. However, unlike F1, F2, and F7, each of the negative features has a moderate contribution to the final decision. The remaining features F4, F5, and F11 are shown to have marginal contributions to the model's decision for this case, and F9 was ranked as the least important feature. In summary, with strong positive attributions from F1, F2, F7, and F4, the model is very certain about the classification verdict, with a certainty of 100.0%.

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

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

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

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

Low

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

KNeighborsClassifier

C1

Company Bankruptcy Prediction

For the case under consideration, the model's output labelling decision is as follows: there is no possibility that C2 is the label for the given case, C1 is the most likely class label, with a confidence level close of 100.0%. The values of the input features, F7, F89, F12, F47, F88, F57, and F51, are the main driving forces resulting in the above classification. The features with moderate influence on the decision here are F17, F50, F74, F19, F49, F55, F28, F37, F60, F54, F24, F48, and F84. Apart from all the abovementioned input features, all the remaining ones, such as F91, F41, F66, and F70, are shown to be irrelevant to the decision made here. Also per the attribution analysis, not all the influential features support labelling the given case as C1, and these are referred to as negative features since they reduce the probability that C1 is the right label here and these are F51, F50, F24, F48, and F84. The notable positive features increasing the probability that C1 is the right label are F7, F89, F12, and F47.

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

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

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

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

No

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

KNeighborsClassifier

C1

Wine Quality Prediction

Based on the influence of features such as F6, F10, F9, and F1, the classifier is pretty confident that the correct label for the given data is C1, whilst, there is a 10.0% probability that the proper label could be C2. The majority of the features have positive contributions, while only F1, F2, and F5 are the negative features, decreasing the classifier's response towards choosing C1. The notal positive features that increase the classifier's response higher towards label C1 instead of C2 include F6, F10, F4, F3, F11, and F9. Taking into consideration the attributions of the input features, we can attribute the classifier's confidence associated with this prediction to the fact that the negative features only have a moderate impact on the classifier's decision for the given data.

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

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

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

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

high quality

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

LogisticRegression

C2

Music Concert Attendance

The model's prediction for this test case is C2 with an almost 100% confidence level which implies that the likelihood of it being a different class label is closer to 0%. Among the top influential feature-set, F9 has a value shifting the label choice in favour of C1, while the others, F13, F3, and F7, all have a positive impact supporting the decision made by the model to assign the label C2. Other features with positive support or impact on the prediction made include F17, F14, F6, and F2. However, F15, F10, F1, and F4 are the other negatives shifting the prediction decision in the direction of the alternative class label. TO sum up, the positive features clearly outweigh the negative features interms of their contributions, hence the confidence level in the classification output.

{'C2': '98.44%', 'C1': '1.56%'}

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

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

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

< 10k

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

MLPClassifier

C2

Ethereum Fraud Detection

The classification verdict for the selected case is C2, and the model is very certain about that considering the prediction probabilities across the possible classes. The top variables influencing this decision are F10, F22, F12, F17, and F24. Other variables that are regarded as somewhat important are F28, F13, F32, F6, F4, F35, F38, F36, F20, F29, F7, F15, F21, F26, and F25. Among the top variables, F10 and F22 decrease the prediction response; therefore, they are pushing the verdict toward C1. Similar to these features, F28, F13, and F4 negatively support assigning C2 to the case. Positively supporting the predicted label are the features F12, F17, F24, and F32. Unlike all the features mentioned above, the values of the remaining features such as F9, F23, F8, and F5, are unessential when determining the correct label for this case.

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

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

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

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

Not Fraud

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

LogisticRegression

C1

Employee Promotion Prediction

Classifying the given case based on the values of its features, C1 is the best label for the given case since its prediction probability is 99.45%, while C2's is just 0.55 percent. The most relevant factors for the classification or prediction declaration above are F4, F11, and F2, whereas the least influential factors are F3, F7, F1, and F9. The other factors' influence can be described as modest and after further inspecting the direction of effect of the factors, F4, F11, F5, F1, and F9 all contribute positively to giving the label C1. These are the favourable factors that raise the likelihood of C1 being the correct designation, however, F2, F6, and F10 are mostly responsible for minimising the chances of C1 and promoting C2.

{'C2': '0.55%', 'C1': '99.45%'}

{'F4': 'avg_training_score', 'F11': 'KPIs_met >80%', 'F2': 'department', 'F6': 'age', 'F5': 'no_of_trainings', 'F10': 'recruitment_channel', 'F8': 'previous_year_rating', 'F3': 'length_of_service', 'F7': 'education', 'F1': 'region', 'F9': 'gender'}

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

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

Promote

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

LogisticRegression

C1

Concrete Strength Classification

Per the predicted likelihoods across the classes, the model predicts label C1 in this case with a high confidence level. Features F7, F5, F4, and F6 are all driving the model towards the C1 classification, with feature F7 being the strongest driver and F6 being the weak driver among the above mentioned set of features. Features F3 and F8 have moderate negative impact on the C1 classification, while feature F1 has a strong positive impact. Finally, feature F2 has a very weak negative impact on the C1 classification decision driving the model towards assigning C2 to the case here.

{'C1': '90.65%', 'C2': '9.35%'}

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

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

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

Weak

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

KNeighborsClassifier

C1

E-Commerce Shipping

The classifier is very uncertain about the correct class for this example and this is because both classes are shown to be equally likely. The above prediction conclusion is mainly based on the influence of the top input features F5, F6, and F3, while F1, F9, and F4 have less influence on the classifier when classifying the given case. When the direction of influence or contribution of each input feature is examined, only F6, F6, F1, and F4 are revealed to have a positive contribution, improving the classifier's affinity to produce the label C1. The remaining features, F3, F5, F2, F8, F10, and F9 have a negative influence and contribution to the final decision.

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

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

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

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

On-time

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

LogisticRegression

C2

Airline Passenger Satisfaction

C2 is the label assigned to this data instance based on the fact that C1 is shown to be very unlikely, with a prediction probability of only 0.68%. The variables most relevant to increasing the probability of the prediction here are F19, F10, F8, and F20. Other positive features that increase the chances of predicting C2 are F5, F2, and F13, however, unlike F8, F10, F19, and F20, these have only moderate contributions to the model's classification decision for this instance. In contrast, F17 is the only top-ranked feature that led the model to classify towards C1, while other negative features with a moderately low contribution included F18, F14, F7, and F15. The least relevant features are F16, F11, F9, and F3, with a very low influence on the C2 prediction, however, unlike these features, F1 and F22 are shown to have no impact, since their attributions are very close to zero, when determining the correct label for the case under consideration. Finally, F22 and F1, according to the attribution analysis have no impact on the classification decision here.

{'C2': '99.32%', 'C1': '0.68%'}

{'F8': 'Type of Travel', 'F17': 'Customer Type', 'F10': 'Inflight entertainment', 'F20': 'Inflight wifi service', 'F19': 'Departure\\/Arrival time convenient', 'F7': 'Gate location', 'F15': 'Arrival Delay in Minutes', 'F18': 'Seat comfort', 'F14': 'Online boarding', 'F5': 'Ease of Online booking', 'F2': 'Class', 'F13': 'Age', 'F12': 'On-board service', 'F4': 'Cleanliness', 'F21': 'Checkin service', 'F6': 'Inflight service', 'F16': 'Food and drink', 'F11': 'Departure Delay in Minutes', 'F9': 'Baggage handling', 'F3': 'Gender', 'F1': 'Flight Distance', 'F22': 'Leg room service'}

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

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

neutral or dissatisfied

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

BernoulliNB

C2

Customer Churn Modelling

C2 is the class assigned to this case or instance. However, according to the classifier, there is a 5.75% chance that the other label, C1, is the correct one. The labelling decision above is mainly due to the values F5, F10, and F7. F4 and F8 are the least ranked features since they have marginal attributions. F10, F3, F6, and F5 have values, increasing the odds of C2 being the correct label and these four features are commonly known as positive variables given that they support the classifier's output decision for the given case. The remaining variables had negative attributions, driving the classification decision towards label C1 and the most negative variables are F7, F2, and F9.

{'C2': '94.25%', 'C1': '5.75%'}

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

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

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

Stay

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

LogisticRegression

C1

Real Estate Investment

The model predicts the class label of this test case or instance as C1 and it is quite confident in the above prediction decision considering the predicted confidence level. The above prediction decision was made primarily based on the values of the following features: F14, F20, F9, and F12. The top features, F14 and F20, positively contribute to the final prediction of C1. Besides, F12 also has a positive impact, pushing the model to output C1. However, the value of F9 supports the prediction of the alternative label, C2. However, compared to F14 and F20, the influence of F9 is very small. The features with moderate influence or impact on the prediction made for this test case are F5, F7, and F13. While F5 moderately supports the C1 prediction, F7 and F13 have values, pushing the model toward predicting C2.

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

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

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

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

Invest

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

DecisionTreeClassifier

C2

Concrete Strength Classification

The case is labelled as C2 by the classification model, and according to the model, there is little to no chance that the correct label could be C1. Per the feature attribution inspection, F2 and F7 are the least influential features. The classification decision to label this case as C2 is mainly due to the positive contributions of F5, F3, and F6. However, the strong negative influence of F1 indicates that the true label could be C1, but since the likelihood of C1 is 0.0%, we can say that the positive features successfully drive the decision in favour of the C2 label. F8, F4, and F7 are the other negative features that unsuccessfully attempt to shift the decision in favour of C1. From the attribution analysis and the predicted likelihoods across the classes, we can conclude that the model is certain that C1 is not the true label.

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

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

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

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

Strong

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

SVC

C2

Real Estate Investment

The decision of the classification model on the true label with respect to the given case is based on the information provided to it. From the prediction probabilities, C2 is selected by the model as the most likely label, with a very high confidence level equal to 97.49%. According to the attributions analysis, the very high confidence in the validity of C2 can be attributed to the very strong positive influence of F19, F4, and F17. The contributions of all the other features are moderate to low. The least relevant features are F13, F1, F16, and F15, whereas the moderate ones include F12, F9, F8, and F18. The very marginal uncertainty with respect to the classification decision here can be blamed on the moderate influence of negative features such as F12, F9, F8, F7, F6, and F18. Aside from F19, F4, and F17, some of the other positive features are F11, F2, and F1, with moderate to low contributions, pushing the decision further higher towards C2 away from C1. Finally, F15 has a negligible contribution to the decision above.

{'C1': '2.51%', 'C2': '97.49%'}

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

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

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

Invest

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

KNeighborsClassifier

C3

Cab Surge Pricing System

With a moderate likelihood of 50.0%, the label for this case is judged to be C3. The classifier, on the other hand, says that C2 and C1 are equally likely, with a predicted probability of 25.0 percent. The aforementioned decision is mostly dependent on the features of the given case and the values of F2, F5, and F1 are demonstrated to be the primary factors influencing the classification output decision. When compared to F2, F5, and F1, the other variables, such as F4, F9, and F8, have lower attributions. According to the attribution assessment, F2, F5, F1, F9, and F7 are the factors that positively contribute to the choice, implying that they are the ones that push the classification closer towards C3. F4, F8, F6, F10, and F11, on the other hand, are the top negative factors that sway the choice somewhat toward the other labels, C2 and C1. In fact, it is because of these negative variables that the classifier presents the probabilities across the C1 and C2.

{'C2': '25.00%', 'C1': '25.00%', 'C3': '50.00%'}

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

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

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

C3

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

LogisticRegression

C2

Music Concert Attendance

With a prediction probability of around 82.06 percent, the algorithm predicts class C2. In the aforementioned prediction judgment, F10, F5, F3, and F19 are all important. The top positively contributing features supporting the C2 prediction are F10, F5, and F19, while F3 is pushing the final prediction away. F11 also has a positive impact on the categorization, but F1 has a negative impact and finally, F7, F17, F16, and F12 have very little influence on the algorithm among the features, when picking the most appropriate label in this case.

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

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

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

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

> 10k

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

LogisticRegression

C2

House Price Classification

For this test case, the model predicts C2 with 99.93% certainty and what this means is that there is only 0.07% chance that C1 could be the right one. The features with the highest impact are F11, F5, F9, and F13, which are all shown to contribute positively to the prediction decision mentioned above. While F4 and F3 support the prediction, F10 is the feature with the strongest negative support for the prediction. Of the features with a small impact, namely F8, F1, F7, F2, F6, and F12, only F1 and F2 negatively support the prediction while the others positively support it.

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

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

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

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

High

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

BernoulliNB

C1

Employee Promotion Prediction

This model trained on eleven attributes predicts class label C1 for this case with a confidence level equal to 54.21%. This suggests that the likelihood of C2 being the correct label is 45.79%. The classification decision above is mainly based on the influence of the features F3, F4, F7, and F2. The most relevant features are the negative features, F3, F4, and F7. These features are regarded as negative features given that their values are shifting the prediction decision in the direction of C2. The positive attributes are F2, F11, F6, F5, and F8, supporting the model's prediction for this case.

{'C1': '54.21%', 'C2': '45.79%'}

{'F3': 'KPIs_met >80%', 'F4': 'previous_year_rating', 'F7': 'avg_training_score', 'F2': 'department', 'F11': 'education', 'F6': 'recruitment_channel', 'F10': 'no_of_trainings', 'F9': 'length_of_service', 'F5': 'region', 'F1': 'age', 'F8': 'gender'}

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

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

Ignore

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

LogisticRegression

C2

Used Cars Price-Range Prediction

With a moderate confidence level of 67.95%, the model predicts C2 for the case under consideration, but it is important to consider the fact that there is a 32.05% chance that C1 could be the correct label instead. The most influential variables resulting in the aforementioned classification decision are F6, F5, and F7. While F6 and F5 have negative contributions towards the C2 prediction; favouring the assignment of C1 instead, F7 is the top positive contributing feature. F3, F4, and F1 had a small positive effect on prediction, whereas F2 had a smaller negative effect. Finally, F9 is the least relevant variable, and therefore, its negative attribution has no significant influence on the model with respect to the given case.

{'C1': '32.05%', 'C2': '67.95%'}

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

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

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

High

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

LogisticRegression

C2

Employee Promotion Prediction

As per the classification algorithm, the most appropriate label for the given case is C2 because its prediction likelihood is 99.45%, whereas that of C1 is only 0.55%. For the classification or prediction assertion above, the most important variables are F6, F10, and F11, while the least influential variables are F9, F3, F1, and F8. Regarding the direction of influence of the variables, the ones with positive contributions to assigning label C2 are F6, F10, F5, F1, and F8 which in fact increase the odds of C2 being the correct label. Finally, decreasing the odds of C2 and supporting C1 are mainly the values of the variables F11, F4, and F7.

{'C1': '0.55%', 'C2': '99.45%'}

{'F6': 'avg_training_score', 'F10': 'KPIs_met >80%', 'F11': 'department', 'F4': 'age', 'F5': 'no_of_trainings', 'F7': 'recruitment_channel', 'F2': 'previous_year_rating', 'F9': 'length_of_service', 'F3': 'education', 'F1': 'region', 'F8': 'gender'}

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

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

Promote

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

SGDClassifier

C1

House Price Classification

C1 is the label predicted by the classification model employed and looking at the prediction probabilities, it valid to concluded that the model is very certain about the selected label. The features considered most relevant by the model for the above decision are F13, F5, F11, and F2, while those with the least consideration are F9, F3, and F10. On the basis of the analysis, majority of the input features positively affirm the prediction for this case; therefore, it is not surprising that the model chose the C1 label and the positive features include F13, F11, F2, F12, F7, F1, and F4. The three negative features that moderately bias the labelling decision towards C2 are F6, F5, and F8.

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

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

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

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

Low

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

LogisticRegression

C1

E-Commerce Shipping

The confidence level for the prediction made for the given case is 71.57%. F6 has a significant impact on the outcome in the negative. The values F3, F10, F7, F9, F2, F8, and F5 all have a positive impact on the results, but they are still less than the effects of F6. The analysis shows that F6 has the highest impact on the model's prediction decision here, it has an overwhelmingly negative effect. F10, F7, F9, and F2 have a positive effect on the model's prediction. Because of the strength of the F6 feature, all other features have little effect on the outcome. In addition, the uncertainty in the prediction could be attributed to the pull of F6, which drives the model to predict an alternative label.

{'C1': '71.57%', 'C2': '28.43%'}

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

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

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

On-time

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

LogisticRegression

C1

E-Commerce Shipping

53.78% and 46.22%, respectively, are the chance or likelihood of any of the classes C1, and C2 being the appropriate label for the case given here. As a result, it's safe to say that C1 is the most likely label for this situation and F10 is identified as the most influential feature whereas F5, F8, and F9 have very low contributions to the decision made by the classification algorithm with respect to the given case. In addition, F1, F6, F2, F3, F4, and F7 have moderate contributions higher than F5, F8, and F9 but lower than F10. Despite the strong positive influence of F10 and F6 supporting the assignment of C1, the negative influence of F1, F2, F3, F7, and F9 shift the classification judgment fairly towards the C2 label which explains the 46.22% likelihood.

{'C2': '46.22%', 'C1': '53.78%'}

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

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

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

Late

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

LogisticRegression

C2

Used Cars Price-Range Prediction

The output decision for the provided data is C2, with a very high confidence level, based on the output prediction probabilities across the two classes since C1 has a probability of around 0.00%. F10, F5, and F7 are the most influential factors in the above-mentioned label assignment, however F3 and F4 are the least influential. The unusually high degree of confidence associated with the classification choice in this case might be attributable to the fact that the bulk of the input variables exhibit attributions that improve the model's responsiveness towards label C2. F2, F1, and F3 have only the negative contributions, attempting to persuade the model to classify this case as C1. To cut a long story short, the joint contribution of the negative variables is quite low in comparison to that of the positive variables, resulting in the model's certainty in the decision above.

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

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

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

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

Low

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