Multi-variate LSTM with attention mechanism for the Indian stock market

Abstract

The advent of attention mechanism has surpassed numerous benchmarks and enabled widespread progress in the realm of natural language processing (NLP). Nevertheless, they have not been adequately leveraged in a time-series context. Accordingly, this paper aims to address this issue by proposing a hybrid, deep-learning model that integrates attention mechanisms and multi-variate long short-term memory (LSTM) for financial forecasting in the Indian stock market. Our model yields superior results as compared to baseline and state-of-the-art models evaluated using MAE and RMSE. Moreover, we employed a modern evaluation criterion based on the methodology advocated by Diebold–Mariano, known as the Diebold–Mariano test (DM test), as a new criterion for evaluation based on statistical hypothesis tests. DM test has been applied in this study to distinguish the significant differences in forecasting accuracy between LSTM with attention and other models. From the results and according to DM-test it is observed that the differences between the forecasting performances of models are significant and that attention mechanism could enhance the accuracy in predicting stock prices by allowing the model to prioritize and concentrate on the most important features and patterns in the data while avoiding overfitting and noise.