Update app.py

app.py

@@ -179,6 +179,20 @@ questions = {
 def predict_ncd(*inputs):
     # Convert inputs to appropriate format
     input_dict = {name: value for name, value in zip(questions.keys(), inputs)}
+    
+    # Encode categorical variables
+    for col, value in input_dict.items():
+        if col in questions and questions[col]['type'] == 'radio':
+            # Find the index of the selected choice
+            try:
+                choice_index = questions[col]['choices'].index(value)
+                # Get the corresponding encoded value
+                input_dict[col] = questions[col]['values'][choice_index]
+            except ValueError:
+                # Handle case where value is not in choices
+                input_dict[col] = -1  # default value for unknown/invalid choices
+    
+    # Create DataFrame with encoded values
     df = pd.DataFrame([input_dict])
     
     # Load the model and scaler
@@ -187,6 +201,15 @@ def predict_ncd(*inputs):
     with open('scaler.pkl', 'rb') as f:
         scaler = pickle.load(f)
     
+    # Ensure column order matches training data
+    expected_columns = ['age', 'A2', 'A1', 'M6', 'M7', 'E5_a', 'education', 'F10', 'E3', 
+                       'B3', 'B4', 'E1', 'E2', 'B1', 'B2', 'E11', 'B5', 'F15', 'B6', 
+                       'J14', 'J15', 'J17', 'K2', 'J1', 'J2', 'J4', 'E12', 'E6', 'B8', 'B9']
+    df = df[expected_columns]
+    
+    # Convert all values to float
+    df = df.astype(float)
+    
     # Preprocess the input
     X = scaler.transform(df)
     
@@ -206,7 +229,7 @@ for name, config in questions.items():
     elif config['type'] == 'radio':
         inputs.append(gr.Radio(choices=config['choices'], label=config['label']))
 
-output = gr.Label(label="NCD Prediction Probabilities")
+output = gr.Label(label="Model Prediction:")
 
 iface = gr.Interface(
     fn=predict_ncd,