app.py

# import gradio as gr
# import pickle
# import pandas as pd 
# import numpy as np
# import lightgbm as lgb
# from autogluon.tabular import TabularPredictor

# loaded_model = pickle.load(open('Autogluon/models/XGBoost_BAG_L1/model.pkl', 'rb'))
# # pred_proba_radio = loaded_model.predict_proba(X_test_radio)
# # pred_proba_radio


# def relapse(age, pathology, B_symptoms):
    
#     X_test_radio = pd.DataFrame.from_dict(
#         {
#             "Age": [age],
#             "pathology_Nodular Sclerosis cHL": [1 if pathology == 'Nodular' else 0],
#             "pathology_others": [1 if pathology == 'Others' else 0],
#             "B_symptoms_Yes" : [1 if B_symptoms else 0],
#             'radio_Yes': [1]
#         }
#     )
    
#     X_test_no_radio = pd.DataFrame.from_dict(
#         {
#             "Age": [age],
#             "pathology_Nodular Sclerosis cHL": [1 if pathology == 'Nodular' else 0],
#             "pathology_others": [1 if pathology == 'Others' else 0],
#             "B_symptoms_Yes" : [1 if B_symptoms else 0],
#             'radio_Yes': [0]
#         }
#     )    

#     pred_proba_radio = loaded_model.predict_proba(X_test_radio)
    
#     pred_proba_radio =  round(np.ndarray.item(pred_proba_radio),2)

#     pred_radio = loaded_model.predict(X_test_radio)
    
#     pred_proba_no_radio = loaded_model.predict_proba(X_test_no_radio)  
    
#     pred_proba_no_radio =  round(np.ndarray.item(pred_proba_no_radio),2)

#     pred_no_radio = loaded_model.predict(X_test_no_radio)
    
#     return {"Radio": pred_proba_radio, "No Radio": pred_proba_no_radio}

# iface = gr.Interface(
    
#     title = 'Should we omit radiotherapy?', 
#                      description = 'This model predicts relapse according to risk factors.', 
#                     fn=relapse, 
#                      inputs= [
#                          gr.Number(label = 'Age', show_label = True),
#         gr.Radio( choices = ['Mixed cellularity', 'Nodular', 'Others'], label = 'Pathology', show_label = True),
#                         gr.Checkbox(label = 'B_symptoms', show_label = True)],
#     outputs = gr.Label(label = 'Has a higher probability of relapse', show_label = True)
#     ,live = True,
#     interpretation="default",
# )

# iface.launch()






import gradio as gr
import pandas as pd 
import numpy as np
from autogluon.tabular import TabularPredictor

predictor_1 = TabularPredictor.load("no_path_PET_1/")
predictor_2 = TabularPredictor.load("no_path_PET_2/")
predictor_3 = TabularPredictor.load("no_path_PET_3/")
predictor_4 = TabularPredictor.load("no_path_PET_4/")

## Import mapping dataset

stage_mapping = pd.read_csv('stage_mapping.csv', index_col = 0)

# stage_mapping

def get_encoding(stage):
    return stage_mapping.loc[stage, 'stage_encoded']   # this is the dataframe for mappings that will be created from the training set

def relapse(age_group, stage):
    
    X_test_radio = pd.DataFrame.from_dict(
        {
            "age_group": ["<=15" if age_group else ">15"],
            'radio': ["Yes"], 
            'stage_encoded' :  get_encoding(stage)
            
        }
    )
        
    X_test_no_radio = pd.DataFrame.from_dict(
        {
            "age_group": ["<=15" if age_group else ">15"],
            'radio': ["No"],
            'stage_encoded' :  get_encoding(stage)

        }
    )    
    
    rad_1 = float(round(predictor_1.predict_proba(X_test_radio).iloc[0,1],3))
    rad_2 = float(round(predictor_2.predict_proba(X_test_radio).iloc[0,1],3))
    rad_3 = float(round(predictor_3.predict_proba(X_test_radio).iloc[0,1],3))
    rad_4 = float(round(predictor_4.predict_proba(X_test_radio).iloc[0,1],3))
    
    pred_proba_radio = float(round(np.mean([rad_1, rad_2, rad_3, rad_4]),3))

    no_rad1 = float(round(predictor_1.predict_proba(X_test_no_radio).iloc[0,1],3)) 
    no_rad2 = float(round(predictor_2.predict_proba(X_test_no_radio).iloc[0,1],3)) 
    no_rad3 = float(round(predictor_3.predict_proba(X_test_no_radio).iloc[0,1],3)) 
    no_rad4 = float(round(predictor_4.predict_proba(X_test_no_radio).iloc[0,1],3)) 
    
    
    pred_proba_no_radio = float(round(np.mean([no_rad1, no_rad2, no_rad3, no_rad4]),3))
    
    
    return {"Radio": pred_proba_radio, "No Radio": pred_proba_no_radio}


iface = gr.Interface(
    
    title = 'Should we omit radiotherapy?', 
                     description = 'This model predicts relapse according to risk factors.', 
                    fn=relapse, 
                     inputs= [
                     gr.Checkbox(label = 'Age <= 15', show_label = True),
                     gr.Dropdown(choices = ['1A', '1B', '2A', '2B', '3A', '3B', '4A', '4B'], 
                                    label = 'stage', show_label = True)                     ]
    ,outputs = gr.Label(label = 'Has a higher probability of relapse', show_label = True)
    ,live = True,
    interpretation="default",
)

iface.launch()