Hybrid Machine Learning and Econometric Models for Predicting BSE-IT Index Prices

Abstract

The forecasting ability of ARIMA, GARCH, XGBoost, and LSTM model is compared in this research in the scenario of forecasting the BSE-IT index stock price. Historical monthly data were trained and tested on these models with significant performance metrics such as MAE, MSE, RMSE, MAPE, and R² score. Findings suggest that XGBoost performs best among other models with the smallest error rates and highest predictive capability. Machine learning algorithms (XGBoost, LSTM) have better flexibility towards stock market patterns than statistical models. The research indicates that hybrid methods for financial prediction are necessary and propose the integration of sentiment analysis and macroeconomic data for increased accuracy.