Research on Influential Factors in Stock Market Prediction with LSTM

Abstract

Stock market prediction is the prediction of stock index or single stock price by time series data prediction algorithms such as machine learning or deep learning. It is used to assist stock market supervision or stock trading, etc. This paper studies the influential factors of stock market prediction by deep learning model long-short term memory (LSTM). Three kinds of influential factors are selected including historical Shanghai A-share index, the historical US NASDAQ index and the term frequency of "increase positions" and "decrease positions" in Weibo. And in order to see their performance, the influential factors are made into two kinds of groups. One is single factor group and the other is multiple factor group. The experiment is carried out on the data during January 1, 2018 to December 31, 2020. And the prediction results are measured by three evaluation metrics: mean absolute error MAE, root mean square error RMSE and mean absolute percentage error MAPE. The experiment results show that the three evaluation metrics of the prediction results of single factor group are better than the prediction results of multiple factor group in both terms of the prediction of stock index value and its daily rise and fall. This paper shows that in the research of stock market index trend prediction, it is not the more predictive factors, the better the prediction results.