Predictive modular fuzzy systems for intelligent sensing

We introduce the predictive modular fuzzy system (PREMOFS) to perform time series classification. A PREMOFS consists of a bank of predictors and a fuzzy inference module. Assuming that the time series is generated by a source belonging to a finite search set, then the classification problem is to select the source that best represents the observed data. The classification is based on a membership function, updated adaptively according to the predictive performance of each model. Two algorithms are presented for updating the membership function: the first one is based on the sum/product fuzzy inference; the second one is based on the max/min fuzzy inference. PREMOFS is a fuzzy modular system which classifies time series to one of a finite number of classes, using the full set of unpreprocessed past data to perform a recursive, adoptive, competitive computation of membership function, based on predictive power. We prove the convergence for both PREMOFS algorithms. We also present simulation results where PREMOFS are applied to signal detection, system identification and phoneme classification tasks