Testing the Suitability of Wavelet Preprocessing for TSK Fuzzy Models

Forecast performance on time serial data by soft computing models like fuzzy systems depends critically, in some cases, on the preprocessing methods used. Time series that exhibit changes in variance require preprocessing, and wavelet-based preprocessing provides a 'natural', parameter-free method for decomposing such time series. However, there are cases where the variance structure of a time series is homogeneous and wavelet-based preprocessing leads to worse results compared to an equivalent analysis carried out using raw data. An automatic method for detecting variance breaks in time series is used as an indicator as to whether or not wavelet-based preprocessing is required. We have evaluated our method by using ten economic time series from the US Census Bureau and Federal Reserve Board, and the results appear to have promise.