Uploaded files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+artifacts/model_training/model.keras filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,36 @@
+import streamlit as st
+import sys
+from pathlib import Path
+import plotly.graph_objs as go
+
+root_dir = Path(__file__).resolve().parents[1]
+sys.path.append(str(root_dir))
+
+import forecaster
+
+def main():
+    st.set_page_config(page_title="Nifty 50 Forecast")
+    st.title("Time Series Forecasting - Nifty 50")
+    
+    if st.button("Forecast"):
+        with st.spinner('Forecasting...'):
+            forecast_df = forecaster.main()
+            forecast_df = forecast_df.round(2)
+            
+            fig = go.Figure()
+            fig.add_trace(go.Scatter(x=forecast_df.index, y=forecast_df["Close"], mode='lines+markers', name='Forecasted Values'))
+            
+            fig.update_layout(
+                title='Nifty 50 Close Forecast',
+                xaxis_title='Date',
+                yaxis_title='Close (Forecasted)',
+                xaxis_tickangle=-90,
+                width=800,
+                height=600
+            )
+            
+            st.plotly_chart(fig)
+            st.write(forecast_df)
+
+if __name__ == "__main__":
+    main()

artifacts/data_preprocessing/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1daff063af1b22ef2d61371bd1be3c1e29ec5828ae59a67f9173db4bdbee8454
+size 1895

artifacts/model_training/model.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c70843511ac13463887b37185236931e83f2923017bdca5e356e7d605e0068e3
+size 2097961

forecaster.py

@@ -0,0 +1,100 @@
+import os
+import joblib
+from matplotlib import pyplot as plt
+import numpy as np
+import pandas as pd
+import tensorflow as tf
+from tsForecaster.components.data_preprocessing import DataPreProcessing
+from tsForecaster.config.configuration import ConfigurationManager
+from tsForecaster.constants import CONFIG_FILE_PATH, PARAMS_FILE_PATH
+from tsForecaster.utils.common import read_yaml
+
+def create_features(df, pred):
+    
+    df = df['Close%']
+    new_date = pd.date_range(df.index[-1] + pd.DateOffset(days=1), periods=1, freq='D')
+    new_close_value = pred
+    
+    new_row = pd.DataFrame({'Close%': new_close_value}, index=[pd.to_datetime(new_date.values[0])])
+    df = pd.concat([df, new_row])
+    
+    df['Close_1D_ago'] = df['Close%'].shift(1)
+    df['Close_2D_ago'] = df['Close%'].shift(2)
+    df['Close_3D_ago'] = df['Close%'].shift(3)
+    df['Close_1W_ago'] = df['Close%'].shift(7)
+    df['Close_2W_ago'] = df['Close%'].shift(14)
+    df['Close_1M_ago'] = df['Close%'].shift(30)
+    df['Close_2M_ago'] = df['Close%'].shift(60)
+    df['Close_3M_ago'] = df['Close%'].shift(90)
+    df['Close_6M_ago'] = df['Close%'].shift(180)
+    df['Close_1Y_ago'] = df['Close%'].shift(365)
+    df['Close_2Y_ago'] = df['Close%'].shift(930)
+    df['Close_3Y_ago'] = df['Close%'].shift(1095)
+    df['Close_5Y_ago'] = df['Close%'].shift(1825)
+    df['Close_7Y_ago'] = df['Close%'].shift(2555)
+    df['Close_10Y_ago'] = df['Close%'].shift(3650)
+    
+    return df
+
+def forecast(model, df, X, time_steps, scaler, close_df, future_steps=15, fig=False) -> pd.DataFrame:
+    future_predictions = []
+    last_row = X[-1:]
+    df_forecast = df.copy()
+
+    for step in range(future_steps):
+        
+        next_close = model.predict(last_row)
+        future_predictions.append(next_close[0, 0])
+        df_forecast = create_features(df_forecast, next_close[0, 0])
+        last_row = df_forecast.iloc[-60:, 1:].values.reshape(1, time_steps, 15)
+
+    predicted_prices = np.array(future_predictions).reshape(-1, 1)
+    predicted_full_features = np.concatenate((predicted_prices, np.zeros((future_steps, 15))), axis=1)
+    predicted_values = scaler.inverse_transform(predicted_full_features)[:, 0]
+    predicted_values_list = predicted_values.flatten().tolist()
+
+    last_close = close_df.values[-1]
+
+    future_closes = []
+
+    for i in range(predicted_values_list.__len__()):
+        next_close = last_close * (1 + (predicted_values_list[i] / 100))
+        future_closes.append(next_close)
+        next_close = last_close
+
+    # Create a DataFrame for future predictions
+    future_dates = pd.date_range(start=df.index[-1] + pd.DateOffset(1), periods=future_steps, freq='D')
+    future_df = pd.DataFrame(future_closes, index=future_dates, columns=['Close'])
+
+    if fig:
+        plt.figure(figsize=(15, 5))
+        plt.plot(future_dates, future_df)
+        plt.show()
+    
+    return future_df
+
+def main() -> pd.DataFrame:
+    config_file = read_yaml(CONFIG_FILE_PATH)
+    params = read_yaml(PARAMS_FILE_PATH)
+    
+    config = ConfigurationManager()
+    data_ingestion_config = config.get_data_ingestion_config()
+    data_preprocessing = DataPreProcessing(config=data_ingestion_config)
+    
+    df = data_preprocessing.process_csv()
+    close_df = df['Close']
+    df.drop(columns=['Close'], inplace=True)
+
+    scaled_df = data_preprocessing.scaling_data(df)
+    scaler = joblib.load(os.path.join(config_file.data_ingestion.scaler_path, 'scaler.pkl'))
+    
+    X, y, dates = data_preprocessing.create_sequences(scaled_df, params.TIME_STEPS)
+    model = tf.keras.models.load_model(config_file.model_training.trained_model_path)
+    
+    forecast_df = forecast(model, scaled_df, X, params.TIME_STEPS, scaler, close_df, future_steps=15)
+    
+    return forecast_df
+
+if __name__ == '__main__':
+    forecast = main()
+    print(forecast)

requirements.txt

@@ -0,0 +1,18 @@
+tensorflow==2.17.0
+opendatasets==0.1.22
+yfinance==0.2.40
+pandas==2.2.2
+numpy
+matplotlib==3.9.1
+seaborn==0.13.2
+python-box==7.2.0
+pyYAML==6.0.1
+tqdm==4.66.4
+ensure==1.0.4
+joblib==1.4.2
+types-PyYAML==6.0.12
+fastapi==0.111.0
+streamlit==1.36.0
+plotly==5.22.0
+scikit-learn
+-e .