Forecasting Fuzzy Time Series with Different Degrees of Membership by Neural Networks

• Activity Number: 523
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Memmedaga Memmedli and Ozer Ozdemir*+
• Companies: Anadolu University and Anadolu University
• Keywords: Forecasting ; Fuzzy time series ; Degrees of membership ; Neural networks ; Fuzzy sets ; Nonlinear

Abstract

Neural network have been successfully used in nonlinear problems and relationships. To forecast fuzzy time series, this study aims to improve forecasting performance with different degrees of membership in establishing fuzzy relationships by using neural networks. Hence, this study applies back-propagation neural networks to establish the fuzzy relationships with different number of hidden nodes. In addition, different lengths of intervals and various neural network architectures are used to improve forecasting. A well-known time series which are enrollment data for the University of Alabama is chosen to compare with other methods proposed in the literature. It is found that the obtained results with this method outperform the other methods.