Upload folder using huggingface_hub

app.py

@@ -28,25 +28,43 @@ def home():
 def predict_sales():
     try:
         data = request.get_json()
-        # Convert input to DataFrame
-        df_input = pd.DataFrame([data])
         
-        # Apply One-Hot Encoding to match training preprocessing
-        categorical_cols = ['current_occupation', 'first_interaction', 'profile_completed', 
-                           'last_activity', 'print_media_type1', 'print_media_type2', 
-                           'digital_media', 'educational_channels', 'referral']
-        
-        df_encoded = pd.get_dummies(df_input, columns=categorical_cols, dtype=int)
-        
-        # Ensure all columns from training are present (add missing as 0)
-        for col in EXPECTED_FEATURES:
-            if col not in df_encoded.columns:
-                df_encoded[col] = 0
-        
-        # Reorder columns to match training set exactly
-        df_encoded = df_encoded[EXPECTED_FEATURES]
+        # Create a dictionary to hold the processed features
+        processed_data = {}
+
+        # 1. Map Numerical Features
+        processed_data['age'] = float(data.get('age', 0))
+        processed_data['website_visits'] = float(data.get('website_visits', 0))
+        processed_data['time_spent_on_website'] = float(data.get('time_spent_on_website', 0))
+        processed_data['page_views_per_visit'] = float(data.get('page_views_per_visit', 0))
+
+        # 2. Map Categorical Features (One-Hot Encoding logic)
+        # Note: 'Professional', 'Mobile App', 'High', 'Email Activity', and 'No' are the baselines (drop_first)
         
-        prediction = model.predict(df_encoded).tolist()[0]
+        occ = data.get('current_occupation', '')
+        processed_data['current_occupation_Student'] = 1 if occ == 'Student' else 0
+        processed_data['current_occupation_Unemployed'] = 1 if occ == 'Unemployed' else 0
+
+        processed_data['first_interaction_Website'] = 1 if data.get('first_interaction', '') == 'Website' else 0
+
+        prof = data.get('profile_completed', '')
+        processed_data['profile_completed_Low'] = 1 if prof == 'Low' else 0
+        processed_data['profile_completed_Medium'] = 1 if prof == 'Medium' else 0
+
+        last = data.get('last_activity', '')
+        processed_data['last_activity_Phone Activity'] = 1 if last == 'Phone Activity' else 0
+        processed_data['last_activity_Website Activity'] = 1 if last == 'Website Activity' else 0
+
+        processed_data['print_media_type1_Yes'] = 1 if data.get('print_media_type1', '') == 'Yes' else 0
+        processed_data['print_media_type2_Yes'] = 1 if data.get('print_media_type2', '') == 'Yes' else 0
+        processed_data['digital_media_Yes'] = 1 if data.get('digital_media', '') == 'Yes' else 0
+        processed_data['educational_channels_Yes'] = 1 if data.get('educational_channels', '') == 'Yes' else 0
+        processed_data['referral_Yes'] = 1 if data.get('referral', '') == 'Yes' else 0
+
+        # Convert dictionary to DataFrame with the exact training order
+        df_final = pd.DataFrame([processed_data])[EXPECTED_FEATURES]
+
+        prediction = model.predict(df_final).tolist()[0]
         return jsonify({'Sales': float(prediction)})
     except Exception as e:
         return jsonify({'error': str(e)}), 400