Predicting movement directions of stock index futures by support vector models with data preprocessing

Abstract

Due to the nature of high-leverage, generous remuneration can be earned by small capital investment. Therefore, analysis of futures prices becomes one of the most interesting topics in financial markets. Recently, by applying the structure risk minimization principle, support vector machines (SVM) approach has been one of the most power techniques to dealing with classification problems. In this investigation, trading information including technical indicators is employed by SVM model to predict movement directions of Taiwan stock index futures prices. Due to data preprocess has essential influence on prediction accuracy of SVM models, preprocessed data provides by different methods are used to examine impacts on prediction performance of SVM models. Experimental results reveal that the SVM approach has the best performance when data are processed by scaling and differencing operations.