A long short term memory and a discrete wavelet transform to predict the stock price

Abstract

Financial Analysis is a challenging task in the present-day world, where investment value and quality are paramount. This research work introduces the use of a prediction technique that uses a combination of Discrete Wavelet Transform (DWT) and Long Short-Term Memory (LSTM) to predict stock prices in the Saudi stock market for the subsequent seven days. A time series model is used where comprises the historical closing values of several stocks listed on the Saudi stock exchange. This model is called the Discrete Long Short-Term Memory (DLSTM) which comprises memory elements that preserve data for extended periods. The function determined the historical closing price of the stock market and then employed Autoregressive Integrated Moving Average (ARIMA) for analysis. The DLSTM-based experimental model had a prediction accuracy of 97.54%, while that of ARIMA was 97.29%. The results indicate that DLSTM is an effective tool for predicting the prices in the stock market. The results highlight the importance of deep learning and the concurrent use of several information sources to predict stock price levels.