Comparative Study on Stock Price Forecasting Based on Deep Learning

Abstract

Stock market forecasting has been a classic but challenging problem due to the attention of economists and computer scientists, with deep learning models being introduced as a new frontier. Stock price data has the characteristics of time series. Based on machine learning long short-term memory (LSTM) that has the advantage of analyzing the relationship between time series data, this paper proposes a stock price prediction method based on the ensemble model of LSTM in an innovative manner. At the same time, prediction models such as CNN, LSTM, Bi-LSTM, GRU, and ensemble LSTM are used to predict stock prices. Results show that the ensemble model of LSTM can provide reliable stock price prediction with the highest prediction accuracy. This forecasting method provides a new sight for stock price forecasting and injects practical experience for scholars to study financial time series data.