Water Quality Prediction Based on a Novel Fuzzy Time Series Model and Automatic Clustering Techniques

Abstract

In recent years, Fuzzy time series models have been widely used to handle forecasting problems, such as forecasting exchange rates, stock index and the university enrollments. In this paper, we present a novel method for fuzzy forecasting designed with the use of the two key techniques, namely automatic clustering and the probabilities of trends of fuzzy trend logical relationships. The proposed method mainly utilizes the automatic clustering algorithm to partition the universe of discourse into different lengths of intervals, and calculates the probabilities of trends of fuzzy trend logical relationships. Finally, it performs the forecasting based on the probabilities that were obtained in the previous stages. We apply the presented method for forecasting the water temperature and potential of hydrogen of the Shing Mun River, Hong Kong. The experimental results show that the proposed method outperforms Chen’s and the conventional methods.