Enhancing stock market predictions using hybrid machine learning approach with XGBOOST-LSTM and XGBOOST-GRU models

In the ever uncertain and volatile global world, it is a trend to extract valuable insights to inform decisions within investment sector. Predicting the performance of stock indices remain a challenge, necessitating continuous data analysis with integration models. Recent application of hybrid ML-models uncovers it is potentiality to construct and work on such forecasting model. Also, the stock market periodically faces some of the unpredictable events like global financial crisis, COVID-19, Russia-Ukraine war, etc., which have created need to explore the hidden variations in stock index movements. To address this need, we use a machine learning approach that achieves predictive performance comparable to other established methods. We propose and evaluate various ML models, specifically the hybrid XGBOOST-LSTM and XGBOOST-GRU models, on pre- and post-COVID Bank Nifty Index data to identify gaps in our understanding of market movements and provide a more precise stock market forecast. It utilises XGBOOST to select an optimised set of features, which used to train the LSTM and GRU models. We discover models provided insights into index data movements, demonstrating superior accuracy and predictive capabilities of the proposed hybrid models.