Update app.py

app.py

@@ -4,83 +4,60 @@ import numpy as np
 import plotly.graph_objects as go
 from datetime import datetime, timedelta
 import pickle
-import os
-import warnings
-warnings.filterwarnings('ignore')
-
-# TensorFlow/Keras imports
-from tensorflow.keras.models import load_model
-from sklearn.preprocessing import MinMaxScaler
-
-# ARIMA and Prophet
+import yfinance as yf
 from statsmodels.tsa.arima.model import ARIMA
 from prophet import Prophet
+from tensorflow import keras
+from sklearn.preprocessing import MinMaxScaler
+import warnings
+warnings.filterwarnings('ignore')
 
-# --------------------------
-# Load models safely
-# --------------------------
+# Load your saved models (update paths as needed)
+# For Hugging Face, these will be in the same directory as app.py
 def load_models():
+    """Load all three models"""
     try:
-        # ARIMA
+        # Load ARIMA model
         with open('arima_model.pkl', 'rb') as f:
             arima_model = pickle.load(f)
-
-        # Prophet
+        
+        # Load Prophet model
         with open('prophet_model.pkl', 'rb') as f:
             prophet_model = pickle.load(f)
-
-        # LSTM model
-        lstm_model = load_model('lstm_model.keras')
-
-        # LSTM scaler
+        
+        # Load LSTM model and scaler
+        lstm_model = keras.models.load_model('lstm_model.h5')
         with open('lstm_scaler.pkl', 'rb') as f:
             scaler = pickle.load(f)
-
+        
         return arima_model, prophet_model, lstm_model, scaler
     except Exception as e:
         print(f"Error loading models: {e}")
         return None, None, None, None
 
-
+# Global variables for models
 arima_model, prophet_model, lstm_model, scaler = load_models()
-SEQ_LENGTH = 60
+SEQ_LENGTH = 60  # Should match your training
 
-# --------------------------
-# Fetch stock data
-# --------------------------
 def fetch_stock_data(ticker, days=365):
-    """
-    Fetch stock data from local CSV fallback.
-    Community Spaces cannot access the internet.
-    """
-    ticker = ticker.upper().strip()
-    filename = f"{ticker}.csv"
-    if not os.path.exists(filename):
-        return None, f"No data found for {ticker}. Upload {ticker}.csv in the Space root."
-
-    df = pd.read_csv(filename, index_col=0, parse_dates=True)
-    if 'Close' in df.columns:
+    """Fetch stock data from Yahoo Finance"""
+    try:
+        end_date = datetime.now()
+        start_date = end_date - timedelta(days=days)
+        df = yf.download(ticker, start=start_date, end=end_date, progress=False)
+        if df.empty:
+            return None, f"No data found for ticker: {ticker}"
         df = df[['Close']].copy()
-    else:
         df.columns = ['Price']
+        return df, None
+    except Exception as e:
+        return None, str(e)
 
-    df.columns = ['Price']
-    df['Price'] = pd.to_numeric(df['Price'], errors='coerce')
-    df = df.dropna()
-    df = df.tail(days)
-
-    if df.empty:
-        return None, f"No valid data found in {filename} for {ticker}."
-    return df, None
-
-# --------------------------
-# Forecasting functions
-# --------------------------
 def make_arima_forecast(data, days):
+    """Make ARIMA forecast"""
     try:
-        data['Price'] = pd.to_numeric(data['Price'], errors='coerce')
-        data = data.dropna()
-        model = ARIMA(data['Price'], order=(1,1,1))
+        # Retrain ARIMA with recent data (or use loaded model)
+        model = ARIMA(data['Price'], order=(1, 1, 1))
         fitted = model.fit()
         forecast = fitted.forecast(steps=days)
         return forecast.values
@@ -89,8 +66,15 @@ def make_arima_forecast(data, days):
         return None
 
 def make_prophet_forecast(data, days):
+    """Make Prophet forecast"""
     try:
-        prophet_data = pd.DataFrame({'ds': data.index, 'y': data['Price'].values})
+        # Prepare data for Prophet
+        prophet_data = pd.DataFrame({
+            'ds': data.index,
+            'y': data['Price'].values
+        })
+        
+        # Create and fit model
         model = Prophet(
             daily_seasonality=True,
             weekly_seasonality=True,
@@ -98,6 +82,8 @@ def make_prophet_forecast(data, days):
             changepoint_prior_scale=0.05
         )
         model.fit(prophet_data)
+        
+        # Make forecast
         future = model.make_future_dataframe(periods=days)
         forecast = model.predict(future)
         return forecast['yhat'].tail(days).values
@@ -106,28 +92,38 @@ def make_prophet_forecast(data, days):
         return None
 
 def make_lstm_forecast(data, days, model, scaler, seq_length=60):
+    """Make LSTM forecast"""
     try:
+        # Scale the data
         scaled_data = scaler.transform(data[['Price']])
+        
+        # Prepare the last sequence
         last_sequence = scaled_data[-seq_length:].reshape(1, seq_length, 1)
-
+        
         predictions = []
         current_sequence = last_sequence.copy()
+        
+        # Generate predictions day by day
         for _ in range(days):
             pred = model.predict(current_sequence, verbose=0)
-            predictions.append(pred[0,0])
-            current_sequence = np.append(current_sequence[:,1:,:], pred.reshape(1,1,1), axis=1)
-
-        predictions = scaler.inverse_transform(np.array(predictions).reshape(-1,1))
+            predictions.append(pred[0, 0])
+            
+            # Update sequence
+            current_sequence = np.append(current_sequence[:, 1:, :], 
+                                        pred.reshape(1, 1, 1), axis=1)
+        
+        # Inverse transform predictions
+        predictions = scaler.inverse_transform(np.array(predictions).reshape(-1, 1))
         return predictions.flatten()
     except Exception as e:
         print(f"LSTM Error: {e}")
         return None
 
-# --------------------------
-# Plot function
-# --------------------------
 def create_forecast_plot(historical_data, forecasts, ticker, model_names):
+    """Create interactive plotly chart"""
     fig = go.Figure()
+    
+    # Historical data
     fig.add_trace(go.Scatter(
         x=historical_data.index,
         y=historical_data['Price'],
@@ -135,13 +131,13 @@ def create_forecast_plot(historical_data, forecasts, ticker, model_names):
         name='Historical Price',
         line=dict(color='blue', width=2)
     ))
-
-    last_date = pd.to_datetime(historical_data.index[-1])
-    future_dates = pd.date_range(
-        start=last_date + timedelta(days=1),
-        periods=len(forecasts[0])
-    )
-
+    
+    # Generate future dates
+    last_date = historical_data.index[-1]
+    future_dates = pd.date_range(start=last_date + timedelta(days=1), 
+                                  periods=len(forecasts[0]))
+    
+    # Plot forecasts
     colors = ['red', 'purple', 'orange']
     for i, (forecast, name) in enumerate(zip(forecasts, model_names)):
         if forecast is not None:
@@ -153,7 +149,7 @@ def create_forecast_plot(historical_data, forecasts, ticker, model_names):
                 line=dict(color=colors[i], width=2, dash='dash'),
                 marker=dict(size=6)
             ))
-
+    
     fig.update_layout(
         title=f'{ticker} Stock Price Forecast',
         xaxis_title='Date',
@@ -161,88 +157,169 @@ def create_forecast_plot(historical_data, forecasts, ticker, model_names):
         hovermode='x unified',
         template='plotly_white',
         height=600,
-        showlegend=True
+        showlegend=True,
+        legend=dict(
+            yanchor="top",
+            y=0.99,
+            xanchor="left",
+            x=0.01
+        )
     )
+    
     return fig
 
-# --------------------------
-# Main prediction function
-# --------------------------
 def predict_stock(ticker, forecast_days, model_choice):
+    """Main prediction function"""
+    # Validate inputs
     if not ticker:
         return None, "Please enter a stock ticker symbol", None
-
-    data, error = fetch_stock_data(ticker, days=730)
+    
+    ticker = ticker.upper().strip()
+    
+    # Fetch data
+    data, error = fetch_stock_data(ticker, days=730)  # 2 years of data
     if error:
         return None, f"Error: {error}", None
-
+    
+    # Make forecasts based on model choice
     forecasts = []
     model_names = []
-
+    
     if model_choice in ["All Models", "ARIMA"]:
         arima_forecast = make_arima_forecast(data, forecast_days)
         if arima_forecast is not None:
             forecasts.append(arima_forecast)
             model_names.append("ARIMA")
-
+    
     if model_choice in ["All Models", "Prophet"]:
         prophet_forecast = make_prophet_forecast(data, forecast_days)
         if prophet_forecast is not None:
             forecasts.append(prophet_forecast)
             model_names.append("Prophet")
-
+    
     if model_choice in ["All Models", "LSTM"] and lstm_model is not None:
         lstm_forecast = make_lstm_forecast(data, forecast_days, lstm_model, scaler, SEQ_LENGTH)
         if lstm_forecast is not None:
             forecasts.append(lstm_forecast)
             model_names.append("LSTM")
-
+    
     if not forecasts:
-        return None, "Failed to generate forecasts.", None
-
+        return None, "Failed to generate forecasts. Please try again.", None
+    
+    # Create plot
     fig = create_forecast_plot(data, forecasts, ticker, model_names)
-
-    # Forecast table
+    
+    # Create forecast table
     future_dates = pd.date_range(
-        start=pd.to_datetime(data.index[-1]) + timedelta(days=1),
+        start=data.index[-1] + timedelta(days=1), 
         periods=forecast_days
     )
+    
     forecast_df = pd.DataFrame({'Date': future_dates.strftime('%Y-%m-%d')})
     for forecast, name in zip(forecasts, model_names):
         forecast_df[f'{name} Prediction ($)'] = np.round(forecast, 2)
-
-    # Summary
-    summary = f"📊 **Forecast Summary for {ticker}**\n\n" \
-              f"- Current Price: ${data['Price'].iloc[-1]:.2f}\n" \
-              f"- Forecast Period: {forecast_days} days\n" \
-              f"- Models Used: {', '.join(model_names)}\n\n" \
-              f"**Predicted Price Range (Day {forecast_days}):**"
+    
+    # Summary statistics
+    summary = f"""
+    📊 **Forecast Summary for {ticker}**
+    
+    - Current Price: ${data['Price'].iloc[-1]:.2f}
+    - Forecast Period: {forecast_days} days
+    - Models Used: {', '.join(model_names)}
+    
+    **Predicted Price Range (Day {forecast_days}):**
+    """
+    
     for forecast, name in zip(forecasts, model_names):
         final_price = forecast[-1]
         change = ((final_price - data['Price'].iloc[-1]) / data['Price'].iloc[-1]) * 100
         summary += f"\n- {name}: ${final_price:.2f} ({change:+.2f}%)"
-
+    
     return fig, summary, forecast_df
 
-# --------------------------
-# Gradio Interface
-# --------------------------
+# Create Gradio Interface
 with gr.Blocks(title="Stock Price Forecasting", theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 📈 Stock Price Forecasting App\nPredict future stock prices using ARIMA, Prophet, and LSTM models.\nUpload CSV files in the Space root for offline use.")
+    gr.Markdown(
+        """
+        # 📈 Stock Price Forecasting App
+        
+        Predict future stock prices using ARIMA, Prophet, and LSTM models.
+        Enter a stock ticker symbol and select forecast parameters below.
+        
+        **Note:** Predictions are for educational purposes only. Not financial advice.
+        """
+    )
+    
     with gr.Row():
         with gr.Column(scale=1):
-            ticker_input = gr.Textbox(label="Stock Ticker Symbol", placeholder="e.g., AAPL", value="AAPL")
-            forecast_days = gr.Slider(minimum=1, maximum=90, value=30, step=1, label="Forecast Days")
-            model_choice = gr.Radio(choices=["All Models", "ARIMA", "Prophet", "LSTM"], value="All Models", label="Select Model(s)")
-            predict_btn = gr.Button("🔮 Generate Forecast", variant="primary")
+            ticker_input = gr.Textbox(
+                label="Stock Ticker Symbol",
+                placeholder="e.g., AAPL, GOOGL, TSLA",
+                value="AAPL"
+            )
+            
+            forecast_days = gr.Slider(
+                minimum=1,
+                maximum=90,
+                value=30,
+                step=1,
+                label="Forecast Days"
+            )
+            
+            model_choice = gr.Radio(
+                choices=["All Models", "ARIMA", "Prophet", "LSTM"],
+                value="All Models",
+                label="Select Model(s)"
+            )
+            
+            predict_btn = gr.Button("🔮 Generate Forecast", variant="primary", size="lg")
+        
         with gr.Column(scale=2):
             output_plot = gr.Plot(label="Forecast Visualization")
-    output_summary = gr.Markdown(label="Forecast Summary")
-    output_table = gr.Dataframe(label="Detailed Forecast", interactive=False)
-
-    predict_btn.click(fn=predict_stock, inputs=[ticker_input, forecast_days, model_choice],
-                      outputs=[output_plot, output_summary, output_table])
+    
+    with gr.Row():
+        output_summary = gr.Markdown(label="Forecast Summary")
+    
+    with gr.Row():
+        output_table = gr.Dataframe(
+            label="Detailed Forecast",
+            wrap=True,
+            interactive=False
+        )
+    
+    # Examples
+    gr.Examples(
+        examples=[
+            ["AAPL", 30, "All Models"],
+            ["GOOGL", 14, "Prophet"],
+            ["TSLA", 60, "LSTM"],
+            ["MSFT", 45, "ARIMA"],
+        ],
+        inputs=[ticker_input, forecast_days, model_choice],
+    )
+    
+    # Connect the button to the function
+    predict_btn.click(
+        fn=predict_stock,
+        inputs=[ticker_input, forecast_days, model_choice],
+        outputs=[output_plot, output_summary, output_table]
+    )
+    
+    gr.Markdown(
+        """
+        ---
+        ### 📚 About the Models
+        
+        - **ARIMA**: Statistical model for time series forecasting
+        - **Prophet**: Facebook's forecasting tool, excellent for seasonality
+        - **LSTM**: Deep learning model that captures complex patterns
+        
+        ### ⚠️ Disclaimer
+        This tool is for educational and research purposes only. Stock market predictions are inherently uncertain. 
+        Always conduct thorough research and consult with financial advisors before making investment decisions.
+        """
+    )
 
-# Launch
+# Launch the app
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()