A Temporal Neuro-fuzzy System for Time Series Analysis

N.A. Şişman Yilmaz and F.N. Alpaslan (Turkey)

Keywords

neurofuzzy system, unfoldingintime

Abstract

In this paper, a temporal neuro-fuzzy system is presented which provides an environment that keeps temporal rela tionships between input and output variables. The sys tem is used to forecast the future behavior of time series data. It is based on ANFIS neuro-fuzzy system and named ANFIS unfolded in time. The rule base contains tempo ral TSK(Takagi-Sugeno-Kang) fuzzy rules. In the training phase, a modified back-propagation learning algorithm is used. The model is tested on Gas-furnace data which is a benchmark problem.

Important Links:



Go Back