Prediction Of CSI 300 Index Based On Lasso And Neural Network Model Under Bayesian Framework

The stock market not only reflects the current economic situation of a country or region,but also plays a role in predicting the economic development of a country or region.The main research object of this paper is the Shanghai and Shenzhen 300 composite index(hereinafter referred to as Shanghai and Shenzhen 300).CSI 300 is a comprehensive index compiled by selecting 300 stocks with the largest scale,the best liquidity and the most representative in Shanghai and Shenzhen stock markets,representing more than 70%of the market value of China’s A-share market.Therefore,CSI 300 can provide investors with authoritative A-share market investment direction,facilitate investors to track and manage their portfolio,and ensure the stability,representativeness,and operability of the index.This paper mainly studies the change of the closing price of Shanghai and Shenzhen 300 index.The model is established based on the closing price data of CSI 300 index and its constituent stocks from June 11,2021,to December 10,2021.Firstly,the variables are selected by lasso’s regularization method and lasso’s method under Bayesian framework to simplify the model and screen out the important explanatory variables of CSI 300 index;Secondly,multiple combination models are constructed by combining BP neural network and Bayesian neural network respectively.Then the training and test errors of the two models were calculated.Finally,compare the two models and choose a better one.When it comes to variables selection,the prediction error of Lasso in the training set is very small,which is 1/1000 of that in the test set,and there may be over-fitting.There is little difference between the two neural network models in the error junction of the training set and the test set respectively.The Bayesian model reduces the error range of the test set,and the non-Bayesian combined model performs better on the training set.