Update app.py

app.py

@@ -9,7 +9,13 @@ le_operation_type = joblib.load("le_operation_type.pkl")
 le_operator_id = joblib.load("le_operator_id.pkl")
 
 def predict_end_date(start_date, quantity, operation_code, operator_code, day_of_week):
-    # Recreate engineered features
+    # Convert inputs to correct types
+    quantity = float(quantity)
+    operation_code = int(operation_code)
+    operator_code = int(operator_code)
+    day_of_week = int(day_of_week)
+
+    # Log transform and feature engineering
     quantity_log = np.log1p(quantity)
     quantity_log_squared = quantity_log ** 2
     quantity_log_cubed = quantity_log ** 3
@@ -18,7 +24,7 @@ def predict_end_date(start_date, quantity, operation_code, operator_code, day_of
     quantity_operator_interaction_squared = quantity_operator_interaction ** 2
     operation_operator_interaction = operation_code * operator_code
 
-    # Reconstruct feature vector with same structure as training
+    # Match the model's training feature order
     X = pd.DataFrame([[
         quantity_log,
         quantity_log_squared,
@@ -46,11 +52,11 @@ def predict_end_date(start_date, quantity, operation_code, operator_code, day_of
     # Predict
     pred_days = model.predict(X)[0]
 
-    # Convert predicted days to end date
+    # Convert to date and readable duration
+    from datetime import datetime, timedelta
     start_dt = datetime.strptime(start_date, "%d-%m-%Y")
     end_dt = start_dt + timedelta(days=pred_days)
 
-    # Split days into days + hours + minutes
     total_minutes = pred_days * 24 * 60
     days = int(total_minutes // (24 * 60))
     hours = int((total_minutes % (24 * 60)) // 60)