Application and analysis of forecasting stock price index based on combination of ARIMA model and BP neural network

Abstract

Stock price index is a barometer of the national economy, which often shows strong nonlinearity because of various factors. It is necessary to use nonlinear models to improve the accuracy of prediction. We predict Shanghai Securities Composition stock index with ARIMA-BP neural network method, and compare the accuracy with the result of single ARIMA model and BP neural network method. We find that the prediction accuracy of ARIMA-BP neural network is better than the BP neural network, BP neural network is better than linear model ARIMA, which confirms the change of stock price index is nonlinear.