Author(s)

Qingru Wu, Duqiao Han, Jingyi Li

Corresponding Author

Qingru Wu

ABSTRACT

A stock market index is a reference number compiled by a stock exchange or financial service institution to indicate changes in the stock market. Investors use artificial intelligence to build a model to predict the time series data to judge stock market trends. This paper proposes KPCA-EMD-LSTM-regular further to improve the accuracy of stock index price prediction. This paper chooses the more market-representative CSI 300 stock index data. In selecting indicators, the technical indicators are included in the model's input variables, and the impact of fundamental market indicators and technical indicators on the closing price of stock index futures is comprehensively considered. This paper uses the CSI 300 index data for empirical analysis. First, the kernel principal component analysis (KPCA) method is used to reduce the dimensionality of the data. Then a two-level decomposition model EMD-CEEMDAN model is constructed to decompose the closing price and regularized extended short-term memory network (LSTM-regular) model for prediction. The empirical results show that the hybrid model has a small error in predicting the closing price of stock index futures than the individual LSTM, EMD, and KPCA models and has achieved better prediction accuracy.

KEYWORDS

CSI 300, KPCA (Kernel principal component analysis), EMD (Empirical Mode Decomposition), CEEMDAN, LSTM (long short-term memory)