Forecasting Direction of China Security Index 300 Movement with Least Squares Support Vector Machine

Abstract

Due to the complexity of financial market, it is a challenging task to forecast the direction of stock index movement. An accurate prediction of stock index movement may not only provide reference value for the investors to make effective strategy, but also for policy maker to monitor stock market, especially in the emerging market, such as China. In this paper, we investigate the predictability of Least Square Support Vector Machine (LSSVM) by predicting the daily movement direction of China Security Index 300 (CSI 300). For comparing purpose, another artificial intelligence (AI) model, Probabilistic Neural Network (PNN) and two Discriminant Analysis models are performed. Ten technical indicators are selected as input variables of the models. Experimental results reveal that LSSVM method is very promising for directional forecasting for that it outperforms PNN, Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) in both training accuracy and testing accuracy.