Robust time series forecasting using fuzzy inference systems

This paper aims to develop a framework of fuzzy systems for robust time-series forecasting. An improved fuzzy rule extraction algorithm using data mining concept is employed to make the resulting fuzzy system be more robust with respect to the input noises or outliers. The proposed technique in this paper is examined with comprehensive robustness analysis by a classical benchmark time-series forecasting problem: the Mackey-Glass time series. Results and comparisons show that the method performs favorably in terms of both accuracy and robustness.