Research on Stock Market Index Prediction Based on LSTM Model

Abstract

In this paper, we propose a method based on a long short-term memory network (LSTM) to predict the rise and fall of a stock market index, and we select the data of CSI 300 from 2016 to 2020 as the research object of this paper. In this paper, the prediction problem of the index is transformed into a multi-classification problem by doing a multi-valued quantitative classification of the up and down ranges. As for the input of the model, we take the basic trading information of the stock and a variety of derived technical indicators as the feature for the model's training, and finally, we make the multi-classification prediction of the rise and fall of the stock. The experimental results show that the model achieves a high prediction accuracy of 78% in the case of multiple classifications. In addition, several comparison models are built in this paper to study the effects of different feature engineering methods on prediction accuracy, so there are some innovations in feature construction. It is concluded that filtering the features before dimensionality reduction can achieve better performance results. Finally, we compare the method proposed in this paper with other traditional machine learning models and conclude that the LSTM model in this paper yields better results while using the same training and test sets.