Stock market prediction of Bangladesh using multivariate long short-term memory with sentiment identification

Abstract

The prediction of stock market trends is a challenging task due to its dynamic and volatile nature. Research has shown that predicting the stock market, especially in developing nations like Bangladesh, is challenging due to the presence of multiple external factors in addition to technical ones. To address this, this study proposed a novel dataset that includes not only technical stock market data from 2014 to 2021, but also external factors such as news sentiment and other economic indicators like inflation, gross domestic product (GDP), exchange rate, interest rate, and current balance. The goal is to provide a comprehensive view of the Dhaka Stock Exchange (DSE), the largest stock market in Bangladesh. The main objective of this study is to predict the trend of DSE by taking into account both technical stock market data and relevant external factors, and to compare the predictions made with and without using external factors. The study utilized a multivariate long short-term memory (LSTM) neural network for the stock market trend prediction. The experimental results showed that the use of external factors improved the accuracy of the LSTM-based stock market trend predictions by approximately 24%.