Stock Index Forecasting Based on Hybrid ARIMA and LSSVM Methodology

Abstract

Both theoretical and empirical findings have indicated that integration of different models can be an effective way of improving the forecasting accuracy of time series, especially when there is a big difference between the combined models. Autoregressive Integrated Moving Average (ARIMA) model is one of the most popular linear models for time series forecasting. However, ARIMA model cannot effectively capture nonlinear patterns hidden in a time series. As a nonlinear model, Least Squares Support Vector Machine (LSSVM) can be applied to time series forecasting with a high degree of accuracy. Combining ARIMA model and LSSVM may further improve the prediction performance. It can helps investors making investment decisions to forecast stock index effectively. In this paper, a hybridization of ARIMA and LSSVM is proposed to forecast the daily closing price of SSE 180 stock index. The empirical results indicate that when linear and nonlinear models were hybridized properly, the forecasting performance of the hybrid model proposed in this paper outperforms the ARIMA model, LSSVM model and other hybrid models.