FUZZY PREDICTION OF CHAOTIC TIME SERIES BASED ON FUZZY CLUSTERING

The main purpose of this paper is to study a new method to model and predict a chaotic time series using a fuzzy model. First, the GK fuzzy clustering method is used to confirm the input space of the fuzzy model. The goal is to divide the training patterns into representative groups so that patterns within one cluster are more similar than those belonging to other clusters. Then, the Kalman filtering algorithm with singular value decomposition is applied to estimate the consequent parameters of the fuzzy model in order to avoid error delivery and error accumulation. The effectiveness of the proposed method is evaluated through simulated examples, including Mackey-Glass time series and Lorenz chaotic systems. The results show that the proposed method provides effective and accurate prediction.