Optimization of Interval Length for Neural Network Based Fuzzy Time Series


ÖZDEMİR Ö., MEMMEDLİ M.

4th International Conference on Problems of Cybernetics and Informatics (PCI), Baku, Azerbaijan, 12 - 14 September 2012 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/icpci.2012.6486456
  • City: Baku
  • Country: Azerbaijan
  • Keywords: Fuzzy time series, neural networks, optimization, forecasting, interval length, FORECASTING ENROLLMENTS, MODEL
  • Anadolu University Affiliated: Yes

Abstract

Fuzzy time series models have become important in past decades with neural networks. Hence, this study aims to improve forecasting performance of neural network based fuzzy time series by using an optimization function to interval length which affects forecasting accuracy. So, a new approach for improving forecasting performance of neural network-based fuzzy time series is applied with optimization process. The empirical results show that the model with proposed approach by optimization of interval length outperforms other forecasting models proposed in the literature.