CARRX Model Based on LSSVR Optimized by Adaptive PSO

Abstract

CARRX model measures financial volatility using range. To improve the forecasting ability of CARRX model, a new volatility forecasting method combining least squares support vector regression (LSSVR) with adaptive particle swarm optimization (APSO) is proposed to the traditional CARRX model. The non-parametric CARRX model is constructed by the LSSVR and APSO algorithm is designed to select the optimal parameters of LSSVR (LSSVR-APSO-CARRX). The results of application on China stock market show that the LSSVR-APSO-CARRX model is better than the LSSVR-CARRX and CARRX model in out-of-sample forecasting performance.