Added CF

app.py

@@ -1,5 +1,11 @@
+import os
+import pickle
+import random
+
+import dice_ml
 import gradio as gr
 import pandas as pd
+from sklearn.model_selection import train_test_split
 
 
 def read_csv(dataset):
@@ -8,11 +14,9 @@ def read_csv(dataset):
     return data.head(10)
 
 def train_model(input_df, target, test_size, model_name, features_to_drop):
-    import pickle
 
     from sklearn.compose import ColumnTransformer
     from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
-    from sklearn.model_selection import train_test_split
     from sklearn.pipeline import Pipeline
     from sklearn.preprocessing import OneHotEncoder, StandardScaler
 
@@ -49,9 +53,51 @@ def train_model(input_df, target, test_size, model_name, features_to_drop):
     regr = Pipeline(steps=[('preprocessor', transformations),
                             ('regressor', RandomForestRegressor())])
     model = regr.fit(X_train, y_train)
-    pickle.dump(model, open('./' + model_name + '.pkl', 'wb'))
+    pickle.dump(model, open('./' + model_name + '.pkl', 'wb'))  
     return 'Model Saved'
 
+def generate_cfs_total(input_df_T, target_T, radio_T, predefined_T, custom_T, dropped_features_T, freeze_features_T, model_T):
+    path = './data/' + input_df_T + '.csv'
+    data = pd.read_csv(path)
+    model = pickle.load(open('./' + model_T + '.pkl', 'rb'))
+    data = data.drop(dropped_features_T, axis=1)
+    metrics = ['TSV','TPV','TCV','TSL']
+    metrics.remove(target_T[0])
+
+    data = data.drop(metrics, axis=1)
+    features = data.drop(target_T[0], axis=1).columns.tolist()
+    target = data[target_T[0]]
+    datasetX = data.drop('TSV', axis=1)
+    x_train, x_test, y_train, y_test = train_test_split(datasetX,
+                                                    target,
+                                                    test_size=0.2,
+                                                    random_state=0)
+
+    always_immutable = ['AvgMaxDailyTemp','AvgMinDailyTemp','School','DAY','StartTime']
+    freezed = always_immutable + freeze_features_T + [target_T[0]]
+
+    features_to_vary = data.columns.difference(freezed).to_list()
+
+    d = dice_ml.Data(dataframe=data, continuous_features=features, outcome_name=target_T[0])
+    m = dice_ml.Model(model=model, backend='sklearn', model_type='regressor')
+
+    exp = dice_ml.Dice(d, m, method='genetic')
+
+    if radio_T == 'Predefined':
+        random_index = random.randint(0, len(x_train-2))
+        print(int(predefined_T))
+        query_instances = x_test[random_index:random_index+int(predefined_T)]
+
+    elif radio_T == 'Custom':
+        query_instances = custom_T
+    
+    dice_exp = exp.generate_counterfactuals(query_instances, total_CFs=4, desired_range=[0.0, 2.0], features_to_vary=features_to_vary)
+    return dice_exp.visualize_as_dataframe(show_only_changes=True)
+
+
+def generate_cfs_individual(input_df_I, target_I, radio_I, predefined_I, custom_I, dropped_features_I, freeze_features_I, model_I):
+    pass
+
 with gr.Blocks() as demo:
     with gr.Tab('Dataset'):
         gr.Markdown('Visualize the dataset to apply CFML')
@@ -60,6 +106,7 @@ with gr.Blocks() as demo:
                             'Summer','TotalClothingValue','TotalClothingValue+3Binary'], label='Dataset')
         output_vis = gr.DataFrame()
         button_vis = gr.Button(label="Run")
+
     with gr.Tab('Model'):
         gr.Markdown('Choose the features to apply CFML')
         input_df = gr.Dropdown(['IndividualClothingBinary','IndividualClothingBinary+3Binary',
@@ -73,8 +120,64 @@ with gr.Blocks() as demo:
         model_output = gr.Textbox(label='Status')
         button_model = gr.Button(label="Train Model")
 
+    #list add .pkl files from models folder
+    models = []
+    for file in os.listdir('./'):
+        if file.endswith('.pkl'):
+            models.append(file.split('.')[0])
+
+    with gr.Tab('Counterfactuals-Total'):
+        gr.Markdown('Generate Counterfactuals for Total CLO Dataset')
+        input_df_T = gr.Dropdown(['Multi_TotalCLO_w_Chair','Summer',
+                                  'TotalClothingValue','TotalClothingValue+3Binary'], label='Dataset')
+        target_T = gr.CheckboxGroup(['TSV','TPV','TCV','TSL'], label='Target Metric', info='Please select only one')
+        #target_T_range = gr.Textbox(label='Target Range', placeholder='Enter the target range [start,end]')
+        radio_T = gr.Radio(['Predefined', 'Custom'], label='Type of Input')
+        predefined_T = gr.Number(default=0, label='Number of inputs to provide')
+        custom_T = gr.Dataframe(
+            headers=['DAY','School','SchoolType','StartTime','AvgMaxDailyTemp','AvgMinDailyTemp','AvgIndoorRelativeHumidity',
+                     'IndoorTempDuringSurvey','Grade','Age','Gender','FormalClothing','TotalCLOwithChair'],
+                     row_count=(2, 'dynamic')
+        )
+
+        dropped_features_T = gr.CheckboxGroup(['SwC', 'MC', 'Grade', 'Age', 'Gender'],
+                                            label='Features to Drop', info='Select the features that are dropped from feature set')
+        
+        freeze_features_T = gr.CheckboxGroup(['SchoolType','StartTime','AvgIndoorRelativeHumidity',
+                     'IndoorTempDuringSurvey','Grade','Age','Gender','FormalClothing','TotalCLOwithChair'],
+                     info = 'Select the features to be freezed to generate CFs')
+        model_T = gr.Dropdown(models, label='Model', info='Select the model to generate CFs')
+        button_cf_T = gr.Button(label="Generate CFs")
+
+    with gr.Tab('Counterfactuals-Individual'):
+        gr.Markdown('Generate Counterfactuals for Individual Clothing Dataset')
+        input_df_I = gr.Dropdown(['IndividualClothingBinary','IndividualClothingBinary+3Binary',
+                            'IndividualClothingValue','IndividualClothingValue+3Binary'], label='Dataset')
+        target_I = gr.CheckboxGroup(['TSV','TPV','TCV','TSL'], label='Target Metric', info='Please select only one')
+        radio_I = gr.Radio(['Predefined', 'Custom'], label='Type of Input')
+        predefined_I = gr.Number(default=0, label='Number of inputs to provide')
+        custom_I = gr.Dataframe(
+            headers=['DAY','School','SchoolType','StartTime','AvgMaxDailyTemp','AvgMinDailyTemp','AvgIndoorRelativeHumidity','IndoorTempDuringSurvey',
+                     'Grade','Age','Gender','FormalClothing','Pant','Trackpant','Halfshirt','Blazer','Jacket','Skirt',
+                     'FullShirt','HalfSweater','Tshirt','Socks','Thermal','Vest','FullSweater','SwC','MC'],
+            row_count=(2, 'dynamic')
+        )
+        dropped_features_I = gr.CheckboxGroup(['SwC', 'MC', 'Grade', 'Age', 'Gender'], 
+                                            label='Features to Drop', info='Select the features that are dropped from feature set')
+        
+        freeze_features_I = gr.CheckboxGroup(['SchoolType','StartTime','AvgIndoorRelativeHumidity','IndoorTempDuringSurvey',
+                                            'Grade','Age','Gender', 'FormalClothing','Pant','Trackpant','Halfshirt','Blazer','Jacket','Skirt',
+                                            'FullShirt','HalfSweater','Tshirt','Socks','Thermal','Vest','FullSweater','SwC','MC'],
+                                            info='Select the features to be freezed to generate CFs')
+        
+
+        model_I = gr.Dropdown(models, label='Model', info='Select the model to generate CFs')
+        button_cf_I = gr.Button(label="Generate CFs")
+
     button_vis.click(read_csv, df_vis, outputs=output_vis)
     button_model.click(train_model, [input_df, target, test_size, model_name, features_to_drop], outputs=model_output)
-
+    button_cf_T.click(generate_cfs_total, [input_df_T, target_T, radio_T, predefined_T, 
+                                           custom_T, dropped_features_T, freeze_features_T, model_T],
+                                           outputs=output_vis)
 
 demo.launch()

requirements.txt

@@ -1,3 +1,4 @@
 numpy
 pandas
-scikit-learn
+scikit-learn
+dice-ml