A new fuzzy time series method based on an ARMA-type recurrent Pi-Sigma artificial neural network

Kocak C., Dalar A. Z., Yolcu Ö., Bas E., Egrioglu E.

SOFT COMPUTING, cilt.24, sa.11, ss.8243-8252, 2020 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 24 Sayı: 11
  • Basım Tarihi: 2020
  • Doi Numarası: 10.1007/s00500-019-04506-1
  • Dergi Adı: SOFT COMPUTING
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, INSPEC, zbMATH
  • Sayfa Sayıları: ss.8243-8252
  • Anahtar Kelimeler: Fuzzy time series, Recurrent Pi-Sigma artificial neural network, Particle swarm optimization, ARMA-type fuzzy time series, Forecasting, FORECASTING ENROLLMENTS, LOGICAL RELATIONSHIPS, TEMPERATURE PREDICTION, MODEL
  • Marmara Üniversitesi Adresli: Hayır

Özet

As it known in many studies, the fuzzy time series methods do not need assumptions such as stationary and the linearity required for classical time series approaches, so there is a huge field of study on fuzzy time series methods in the time series literature. Fuzzy time series literature has the studies which use both the various models of artificial neural networks and the different optimization methods of artificial intelligence jointly. In this study, a new fuzzy time series algorithm based on an ARMA-type recurrent Pi-Sigma artificial neural network is introduced. It is expected that the proposed method increases the forecasting performance for many real-life time series because of using more input which is the error term obtained from Pi-Sigma artificial neural network with recurrent structure. Therefore, it can be considered that the proposed method is based on an ARMA-type fuzzy time series forecasting model. In the proposed method, the training of recurrent ARMA-type Pi-Sigma neural network is performed by particle swarm optimization. The proposed method has been applied to a real-data set as well as simulated data sets of a real-life time series, and the obtained results have been compared with some other methods in the literature.