Update app.py

app.py

@@ -86,38 +86,93 @@ def lstm_gru_forecast(data, model_type, steps):
 def ensemble_forecast(predictions_list):
     return pd.DataFrame(predictions_list).mean(axis=0)
 
-# Streamlit App
-st.title("Stock Price Forecasting App")
-
-# Load stock data
-symbol = 'AAPL'  # Replace with the desired stock symbol
-start_date = '2021-01-01'
-end_date = '2022-01-01'
-stock_prices = get_stock_data(symbol, start_date, end_date)
-
-# ARIMA parameters
-arima_order = (3, 0, 0)  # Example: AR component (p) is set to 3, differencing (d) is 0, MA component (q) is 0
-arima_forecast_steps = 30  # Number of steps to forecast (adjust based on your preference)
-
-# LSTM and GRU parameters
-lstm_gru_forecast_steps = 30  # Number of steps to forecast (adjust based on your preference)
-
-# ARIMA Forecast
-arima_predictions = arima_forecast(stock_prices, arima_order, arima_forecast_steps)
+# Function to fit ARIMA model and make predictions
+def arima_forecast(data, p, d, q, steps):
+    # Differencing
+    for i in range(d):
+        data_diff = np.diff(data)
+        data = data_diff
+
+    # Autoregressive (AR) and Moving Average (MA) components
+    ar_coef = np.zeros(p) if p > 0 else []
+    ma_coef = np.zeros(q) if q > 0 else []
+
+    # Initial prediction
+    predictions = list(data[:p])
+
+    # ARIMA forecasting
+    for i in range(len(data) - p):
+        ar_term = sum(ar_coef[j] * data[i + p - j - 1] for j in range(p))
+        ma_term = sum(ma_coef[j] * (data[i + p - j - 1] - predictions[-1]) for j in range(q))
+        next_prediction = data[i + p] + ar_term + ma_term
+        predictions.append(next_prediction)
+
+        # Update coefficients using online learning (optional)
+        if i + p + 1 < len(data):
+            ar_coef = ar_coef + (2.0 / (i + p + 2)) * (data[i + p + 1] - next_prediction) * np.flip(data[i:i + p])
+            ma_coef = ma_coef + (2.0 / (i + p + 2)) * (data[i + p + 1] - next_prediction) * np.flip(predictions[i - q + 1:i + 1])
+
+    # Inverse differencing
+    for i in range(d):
+        predictions = np.cumsum([data[p - 1]] + predictions)
+
+    return predictions[-steps:]
 
-# LSTM Forecast
-lstm_predictions = lstm_gru_forecast(stock_prices, 'lstm', lstm_gru_forecast_steps)
-
-# GRU Forecast
-gru_predictions = lstm_gru_forecast(stock_prices, 'gru', lstm_gru_forecast_steps)
-
-# Ensemble Forecast (Averaging)
-ensemble_predictions = ensemble_forecast([arima_predictions, lstm_predictions, gru_predictions])
-
-# Plotting
-st.write("### Historical Stock Prices and Forecasts")
-st.line_chart(stock_prices)
-st.line_chart(arima_predictions)
-st.line_chart(lstm_predictions)
-st.line_chart(gru_predictions)
-st.line_chart(ensemble_predictions)
+# Streamlit App
+def main():
+    st.title("Stock Price Forecasting App")
+
+    # Load stock data using Streamlit sidebar
+    symbol = st.sidebar.text_input("Enter Stock Symbol", value='AAPL')
+    start_date = st.sidebar.text_input("Enter Start Date", value='2021-01-01')
+    end_date = st.sidebar.text_input("Enter End Date", value='2022-01-01')
+    stock_prices = get_stock_data(symbol, start_date, end_date)
+
+    # ARIMA parameters using Streamlit sliders
+    p = st.sidebar.slider("AR Component (p)", min_value=0, max_value=10, value=3)
+    d = st.sidebar.slider("Differencing (d)", min_value=0, max_value=5, value=0)
+    q = st.sidebar.slider("MA Component (q)", min_value=0, max_value=10, value=0)
+    arima_forecast_steps = st.sidebar.slider("ARIMA Forecast Steps", min_value=1, max_value=100, value=30)
+
+    # LSTM and GRU parameters using Streamlit sliders
+    lstm_gru_forecast_steps = st.sidebar.slider("LSTM/GRU Forecast Steps", min_value=1, max_value=100, value=30)
+
+    # Custom ARIMA Forecast using Streamlit button
+    if st.sidebar.button("Run Custom ARIMA Forecast"):
+        arima_predictions_custom = arima_forecast(stock_prices.values, p, d, q, arima_forecast_steps)
+        arima_predictions_custom = pd.Series(arima_predictions_custom, index=pd.date_range(start=stock_prices.index[-1], periods=arima_forecast_steps + 1, freq=stock_prices.index.freq))
+
+        # Display ARIMA Forecast Plot
+        st.subheader("Custom ARIMA Forecast")
+        st.line_chart(pd.concat([stock_prices, arima_predictions_custom], axis=1).rename(columns={0: "ARIMA Forecast"}))
+
+    # LSTM Forecast using Streamlit button
+    if st.sidebar.button("Run LSTM Forecast"):
+        lstm_predictions = lstm_gru_forecast(stock_prices, 'lstm', lstm_gru_forecast_steps)
+
+        # Display LSTM Forecast Plot
+        st.subheader("LSTM Forecast")
+        st.line_chart(pd.concat([stock_prices, pd.Series(lstm_predictions, index=pd.date_range(start=stock_prices.index[-1], periods=lstm_gru_forecast_steps + 1, freq=stock_prices.index.freq))], axis=1).rename(columns={0: "LSTM Forecast"}))
+
+    # GRU Forecast using Streamlit button
+    if st.sidebar.button("Run GRU Forecast"):
+        gru_predictions = lstm_gru_forecast(stock_prices, 'gru', lstm_gru_forecast_steps)
+
+        # Display GRU Forecast Plot
+        st.subheader("GRU Forecast")
+        st.line_chart(pd.concat([stock_prices, pd.Series(gru_predictions, index=pd.date_range(start=stock_prices.index[-1], periods=lstm_gru_forecast_steps + 1, freq=stock_prices.index.freq))], axis=1).rename(columns={0: "GRU Forecast"}))
+
+    # Ensemble Forecast using Streamlit button
+    if st.sidebar.button("Run Ensemble Forecast"):
+        ensemble_predictions = ensemble_forecast([arima_predictions_custom, lstm_predictions, gru_predictions])
+
+        # Display Ensemble Forecast Plot
+        st.subheader("Ensemble Forecast")
+        st.line_chart(pd.concat([stock_prices, ensemble_predictions], axis=1).rename(columns={0: "Ensemble Forecast"}))
+
+    # Plotting Historical Stock Prices
+    st.subheader("Historical Stock Prices")
+    st.line_chart(stock_prices)
+
+if __name__ == "__main__":
+    main()