Update app.py

app.py

@@ -23,14 +23,16 @@ class PriceScaler:
         self.scaler = MinMaxScaler()
         
     def fit_transform(self, data):
-        # Reshape data to 2D array
-        reshaped_data = np.array(data).reshape(-1, 1)
-        return self.scaler.fit_transform(reshaped_data).ravel()
+        # Ensure data is 2D for fitting
+        data_2d = np.array(data).reshape(-1, 1)
+        # Transform and return 1D array
+        return self.scaler.fit_transform(data_2d).flatten()
         
     def inverse_transform(self, data):
-        # Reshape data to 2D array
-        reshaped_data = np.array(data).reshape(-1, 1)
-        return self.scaler.inverse_transform(reshaped_data).ravel()
+        # Ensure data is 2D for inverse transform
+        data_2d = np.array(data).reshape(-1, 1)
+        # Transform and return 1D array
+        return self.scaler.inverse_transform(data_2d).flatten()
 
 class CryptoPredictor(nn.Module):
     def __init__(self, input_dim: int, hidden_dim: int = 128, num_layers: int = 2, dropout: float = 0.2):
@@ -130,10 +132,11 @@ class CryptoAnalyzer:
         return df.iloc[-days:]
 
     def prepare_data(self, df: pd.DataFrame, lookback: int) -> Tuple[torch.Tensor, torch.Tensor]:
+        # Scale features
         features = df[self.feature_columns].values
         scaled_features = self.scaler.fit_transform(features)
     
-        # Handle close prices
+        # Scale close prices - ensure 1D output
         close_prices = df['Close'].values
         scaled_close = self.price_scaler.fit_transform(close_prices)
     
@@ -143,7 +146,7 @@ class CryptoAnalyzer:
             y.append(scaled_close[i + lookback])
     
         X = torch.FloatTensor(np.array(X)).to(self.device)
-        y = torch.FloatTensor(np.array(y)).to(self.device)
+        y = torch.FloatTensor(np.array(y)).reshape(-1).to(self.device)
     
         return X, y
 
@@ -199,7 +202,7 @@ class CryptoAnalyzer:
         try:
             df = self.get_data(symbol, days)
             X, y = self.prepare_data(df, lookback)
-    
+
             model_path = self.get_model_path(symbol)
             if os.path.exists(model_path):
                 model = CryptoPredictor(X.shape[2]).to(self.device)
@@ -207,19 +210,18 @@ class CryptoAnalyzer:
             else:
                 model = self.train_model(X, y, symbol)
                 joblib.dump(self.scaler, self.get_scaler_path(symbol))
-    
+
             model.eval()
             with torch.no_grad():
                 predictions, confidence = model(X)
-                predictions = predictions.cpu().numpy()
-                confidence = confidence.cpu().numpy()
+                predictions = predictions.cpu().numpy().flatten()  # Ensure 1D
+                confidence = confidence.cpu().numpy().flatten()  # Ensure 1D
             
-                # Ensure predictions are 1D before inverse transform
-                predictions = predictions.reshape(-1)
+                # Inverse transform predictions
                 predictions = self.price_scaler.inverse_transform(predictions)
             
-                # Handle actual values similarly
-                y_np = y.cpu().numpy().reshape(-1)
+                # Inverse transform actual values
+                y_np = y.cpu().numpy().flatten()  # Ensure 1D
                 actual_prices = self.price_scaler.inverse_transform(y_np)
         
             rmse = float(np.sqrt(np.mean((actual_prices - predictions) ** 2)))