HybStock: A Hybrid Deep Learning-Based Model to Predict Stock Market on Dhaka Stock Exchange

Stock Price Prediction has always been a field of attraction and interest for investors and researchers. Since several factors influence the stock price, predicting it is a difficult task, due to the constant fluctuations in stock prices. However, knowing a company's previous performance in the stock market can track the data and provide predictions to stockholders. Deep learning implementation for stock market analysis came out with promising results in the recent past. Numerous studies attempt to forecast the stock price using several models, even using a single CNN or LSTM. To fully utilize each model's benefits in stock prediction and increase accuracy, our approach suggests a hybrid CNN-LSTM model on the Dhaka Stock Exchange, which performs well and has a strong generalization capacity. Here, characteristics from the stock data are extracted using CNN, and the LSTM is then utilized to find patterns and forecast the stock market. The datasets utilized in our research, include the historical prices of four popular companies of the Dhaka Stock Exchange. In the proposed approach, our hybrid model is compared with some of the most popular forecasting models like the Moving Average Method, Linear Regression, K-NN, Auto Arima, LSTM, and ID-CNN. Using the datasets we predicted the price and evaluated the RMSE for each model. The chosen hybrid CNN-LSTM model demonstrates a high level of accuracy with the least error, which can help investors achieve the purpose of expanding return and avoiding risk.