New iteration

app.py

@@ -24,15 +24,24 @@ EXPECTED_COLUMNS.extend([f'games_{i}' for i in games_ids])
 EXPECTED_COLUMNS.extend([f'winrate_{i}' for i in games_ids])
 
 def load_reference_data(player_id):
-    """Load reference prediction data for comparison"""
+    """Load reference prediction data and input features"""
     try:
+        # Read the full CSV to get both input features and prediction
         ref_df = pd.read_csv(f"{player_id}.csv", encoding='utf-8')
-        print("Reference data columns:", ref_df.columns)
-        print("Reference data values:", ref_df.iloc[-1])
-        return ref_df.iloc[-1][f"{player_id}_S34"]
+        
+        # Extract the features
+        features = ref_df[ref_df.index != 'Predicted_Cost']
+        print("\nReference features:")
+        print(features)
+        
+        # Get the prediction from the last row
+        prediction = ref_df.iloc[-1][f"{player_id}_S34"]
+        print(f"Reference prediction: {prediction}")
+        
+        return features, prediction
     except Exception as e:
         print(f"Could not load reference data: {e}")
-        return None
+        return None, None
 
 def prepare_single_player_data(user_id, mmr, comf_1, comf_2, comf_3, comf_4, comf_5):
     """Creates a DataFrame in the expected format for the model"""
@@ -128,11 +137,19 @@ def predict_cost(user_id, mmr, comf_1, comf_2, comf_3, comf_4, comf_5):
     try:
         # Extract player_id for reference data
         player_id = user_id.split("/")[-1] if "/" in user_id else user_id
-        reference_cost = load_reference_data(player_id)
+        reference_features, reference_prediction = load_reference_data(player_id)
         
         # Prepare the player data
         processed_data = prepare_single_player_data(user_id, mmr, comf_1, comf_2, comf_3, comf_4, comf_5)
         
+        print("\nComparing processed data with reference:")
+        if reference_features is not None:
+            for col in processed_data.columns:
+                if col in reference_features.index:
+                    print(f"{col}:")
+                    print(f"  Our value: {processed_data[col].iloc[0]}")
+                    print(f"  Ref value: {reference_features.loc[col, player_id + '_S34']}")
+        
         # Load and use the model
         model_path = Path("model/rd2l_forest.onnx")
         if not model_path.exists():
@@ -140,10 +157,6 @@ def predict_cost(user_id, mmr, comf_1, comf_2, comf_3, comf_4, comf_5):
             
         session = ort.InferenceSession(str(model_path))
         
-        # Debug information
-        print("\nModel Input Information:")
-        print("Processed data shape:", processed_data.shape)
-        
         # Make prediction
         input_name = session.get_inputs()[0].name
         prediction = session.run(None, {input_name: processed_data.values.astype(np.float32)})[0]
@@ -156,11 +169,11 @@ def predict_cost(user_id, mmr, comf_1, comf_2, comf_3, comf_4, comf_5):
         total_winrate = hero_stats.get('total_winrate', 'N/A')
         
         comparison = ""
-        if reference_cost is not None:
-            diff = abs(predicted_cost - reference_cost)
+        if reference_prediction is not None:
+            diff = abs(predicted_cost - reference_prediction)
             comparison = f"""
-Reference Cost: {reference_cost}
-Difference: {diff:.2f} ({(diff/reference_cost*100):.1f}% {'higher' if predicted_cost > reference_cost else 'lower'})"""
+Reference Cost: {reference_prediction}
+Difference: {diff:.2f} ({(diff/reference_prediction*100):.1f}% {'higher' if predicted_cost > reference_prediction else 'lower'})"""
         
         return f"""Predicted Cost: {predicted_cost}{comparison}