Improvement of ANFIS model by developing of novel hybrid learning algorithms for contraction scour modeling

  • Keivan Kaveh
  • Minh Duc Bui
  • Peter Rutschmann


This paper, introduces the development of two new hybrid learning rules on ANFIS technique as an alternative to common hybrid learning rule in prediction of contraction scour depth. In contrast with common hybrid rule which combines the gradient method and the least squares estimate, new hybrid learning rules combine the Levenberg-Marquardt and the gradient methods, as well as the Levenberg-Marquardt method and the least squares estimate. To this aim, MATLAB toolbox is used to build ANFIS models based on common learning rules and FORTRAN programming language is utilized to construct ANFIS models for the proposed hybrid learning algorithms. The results of this study indicate that the proposed methods perform better than common learning rules and decrease significantly the CPU time compared to the Levenberg-Marquardt method.