Prediction and contribution analysis of CSI 300 Stock Index Futures based on KPCA-LightGBM

Abstract

Stock index is a price index compiled to measure the trend of the overall price level of the stock market. It can sensitively reflect the changing situation of the economy of the host country. It is of great significance to study the influencing factors of stock index and make predictions. (Light Gradient Boosting Machine LightGBM is a new type of integrated algorithm based on decision tree. This paper builds a model based on LightGBM algorithm to predict the trend of stock index futures and calculate the influence of each dependent variable on stock price, which can help investors to make investment decisions and has strong theoretical and practical significance. In this paper, based on the previous research on the prediction model and related literature review, a feature system of 5 first-level indicators and 32 second-level indicators was constructed. Kernel principal component analysis (KPCA) was used for dimensionality reduction. Finally, LightGBM algorithm was used to adjust and train the data set after dimensionality reduction. In this paper, KPCA-LightGBM based CSI 300 stock index futures prediction and fitting precision is high, can basically meet the prediction requirements, has practical investment application value, and lays a foundation for future improvement and research.