c-ascending support vector machines for financial time series forecasting

Abstract

This paper proposes a modified version of support vector machines (SVMs), called c-ascending support vector machines (c-ASVMs), to model non-stationary financial time series. c-ASVMS are obtained by a simple modification of the regularized risk function in SVMs whereby the recent /spl epsiv/-insensitive errors are penalized more heavily than the distant /spl epsiv/-insensitive errors. This procedure is based on the prior knowledge that in the non-stationary financial time series, the recent past data could provide more important information than the distant past data. In the experiment, c-ASVMS are tested using three real futures collected from the Chicago Mercantile Market. It is shown that the c-ASVMS with the actually ordered sample data consistently forecast better than the standard SVMs, with the worst performance when the reversely ordered sample data are used. Furthermore, the c-ASVMs use fewer support vectors than those of the standard SVMs, resulting in a sparser representation of solution.