Modified TSK fuzzy system in stock price forecasting

Abstract

This paper proposes a modified version of TSK fuzzy system, called C -ascending TSK fuzzy system, to model stock price. The C -ascending TSK fuzzy system is obtained by a simple modification of the weighted membership function in TSK fuzzy system, where the recent errors are penalized more heavily than the distant errors. This procedure is based on the prior knowledge that in the non-stationary financial time series the dependency between input variables and output variables gradually changes over the time, especially, the recent past data could provide more important information than the distant past data. In the experiment, C -ascending TSK fuzzy system is tested using the Standard&Poor 500 sock index(CME-SP). It is shown that the C -ascending TSK fuzzy system with the actually ordered sample data consistently forecast is better than the standard TSK fuzzy system.