Time series prediction using fuzzy inductive reasoning a case study

This paper presents the application of fuzzy inductive reasoning (FIR) to time-series analysis and forecasting. This methodology had previously been applied to modeling and control of dynamic input-output systems [de Albornoz 96], [Mugica 95], [Nebot 94]. The research effort discussed in this paper presents a first attempt at assessing the suitability of this qualitative modeling methodology for forecasting time-series, i.e., for predicting the future development of signals on the basis of their own past, without identifying the systems that produce these signals. The performance of FIR in time-series forecasting is compared with Box-Jenkins methods and neural networks in a case study.