A deep learning approach for stock market prediction using deep autoencoder and long short-term memory

Abstract

The stock market prediction problems have received increased attention from researchers due to the high stakes involved and the need for better prediction accuracy. We have developed an architecture by combining a deep autoencoder and long short-term memory to give a novel deep learning framework to forecast the stock price. In stock price forecasting, applying a deep autoencoder that extracts deep features is a new concept. The autoencoder denoise the stock data, and the LSTM model stores past information to predict the future stock price. The deep learning framework that we have used comprises multiple stages. The data is fed into the deep autoencoder to generate a noise-free dataset of the stock price. In the next stage, the deep autoencoder's output is provided as input into the LSTM model to predict the price after n days. Our proposed model could overcome the limitations of traditional machine learning models used in financial prediction. We have validated the model 's effectiveness using multiple datasets and compared the performance with existing models in the literature. The results show that the proposed DAE-LSTM model outperforms the current models.