Enhancing predictive modeling for Indian banking stock trends: A fusion of BERT and attention-based BiLSTM approach

Abstract

In the field of finance, deep learning techniques have been extensively researched for predicting stock prices. In this research, we propose a novel approach for predicting stock price movements using a combination of reviews and historical price data for SBI and HDFC stocks. As market volatility is influenced by numerous factors, it is crucial to consider it while predicting stock prices. To capture the interactions between the price and text data effectively, we create a fusion mix and utilize a hybrid information mixing module, designed using BERT and BiLSTM, to extract the multimodal interactions between the time series and semantic features. The proposed model, the hybrid information mixing module, is based on a multilayer perceptron and achieves high accuracy in predicting price fluctuations in highly volatile stock markets. Future research can extend this approach to include additional data sources and explore other deep learning techniques for better performance.