Construction And Prediction Of Stock Market Sentiment Index Based On PCA And NARX Dynamic Neural Network

At present, most researches about the stock market sentiment index use the single emotion variable, but few on the emotion index. The research of dynamic neural network based on the stock market sentiment index is even less. Most sentiment index construction still stay in the basic analysis of time series, which uses traditional regression model and time series model. The research that using dynamic neural network are less than the traditional time series ones.This paper optimized current construction of the sentiment index, analyzed premise situation of constructing BW sentiment index, proposed a new idea according to the extracted common factors. If each agent variable contains three levels of information, as common basic factors, common emotional factors and characterized factors, the first extracted principal component is investor sentiment after exclusion of basic factors.This paper is different from the existing research by a plurality of single emotion variable, constructed sentiment index used principal analysis, and making empirical investigation through NARX neural network. Finally, it compared by the traditional time series ARIMA model. The conclusion is the forecasting results of the Shanghai composite index from NARX neural network is feasible, which is better than the one produced from ARIMA model.