Dhaka Stock Exchange stock price prediction using Machine Learning and Deep Learning Models

Abstract

The stock market is unstable and generally unpredictable, as any one of us could have predicted. Researchers have been experimenting with time-series data to forecast future values for many years, with stock valuation forecasting being the most difficult and lucrative application. Market movement, however, depends on a variety of factors, only a small subset of which can be quantified, including historical stock data, trade volume, and current pricing. This makes predicting stock prices using machine learning difficult and, to some extent, unreliable. With an adequate amount of historical data and variables, mathematical and machine learning algorithms are used to anticipate short-term market movements for a typical, uninteresting market day. This paper proposes several comparative models for stock price prediction using various machine learning algorithms like Bidirectional LSTM, Multi-Head Attention Based LSTM, Prophet, ARIMA etc. The models have been trained using historical data collected from the Dhaka Stock Exchange (DSE) official website. The financial data contains factors like Date, Volume, Open, High, Low Close, and Adj Close prices. The models are evaluated using standard strategic indicators like Mean Squared error (MSE), Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE) and R-Squared. Moreover, in order to thoroughly understand the predictions, we implemented explainable AI models such as LIME. We believe that the information in this article will be useful to stock investors in determining the best times to buy and/or sell stocks on the Dhaka Stock Exchange.