Parameter estimation by anfis where dependent variable has outlier


ERBAY DALKILIÇ T., ŞANLI KULA K., APAYDIN A.

HACETTEPE JOURNAL OF MATHEMATICS AND STATISTICS, cilt.43, sa.2, ss.309-322, 2014 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 43 Sayı: 2
  • Basım Tarihi: 2014
  • Dergi Adı: HACETTEPE JOURNAL OF MATHEMATICS AND STATISTICS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, TR DİZİN (ULAKBİM)
  • Sayfa Sayıları: ss.309-322
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

Regression analysis is investigation the relation between dependent and independent variables. And, the degree and functional shape of this relation is determinate by regression analysis. In case that dependent variable has outlier, the robust regression methods are proposed to make smaller the effect of the outlier on the parameter estimates. In this study, an algorithm has been suggested to define the unknown parameters of regression model, which is based on ANFIS (Adaptive Network based Fuzzy Inference System). The proposed algorithm, expressed the relation between the dependent and independent variables by more than one model and the estimated values are obtained by connected this model via ANFIS. In the solving process, the proposed method is not to be affected the outliers which are to exist in dependent variable. So, to test the activity of the proposed algorithm, estimated values obtained from this algorithm and some robust methods are compared.