Fuzzy time-depending logical relationship groups in fuzzy time series models

Abstract

Fuzzy time series models have many applications in forecasting, especially in the economic forecast. In recent years many works have been completed towards improving accuracy and reducing the amount calculated m fuzzy time series models such as the article by

Chen and Hsu, Huamg, Kuo. 'W'u. A different approach to improve efficiency for time series prediction is to use fuzzy techniques in data mining such as clustering, neural networks, ... to build the model. The most of methods are based on the technique of Chen’s fuzzy logic relationship groups to reduce the amount of computation to just perform arithmetic calculations instead of min-max as in the model of Song-Chissom. Yu (2005) developped the recurrent fuzzy relationships and contnicted the weighted fuzzy time series model.

In this paper, we propose a modified way to define fiizzy time-depending logical relationship groups. Thanks to the new concept of fuzzy relationship groups, the new fuzzy time series model is proposed. Using this model for forecasting fuzzy time scries model, we obtained better results for enrollments and Taiwan stock index.forecasting than Chen and Yu results.