Financial time series analysis using FWNN with robust training algorithm

Finanicial market is characterized with complex, stochastic, nonstationary process and the development of effective models for prediction of a stock price is one of the important problems in finance. For analyzing nonlinear time-series, the importance of nonlinear models, such as neural networks (NNs) and fuzzy systems (FSs), has been increasing in recent years. Combining NNs, FSs and wavelets, FuzzyWavelet Neural Network (FWNN), which has advantages of each systems, was devised. However, when time-series analysis is actually conducted, these time-series data are influenced by disturbance or noise. So in this paper, we introduce FWNN with robust training algorithm which can guarantee the prediction accuracy to some extent even in such a case.