Design of ANFIS Networks Using Hybrid Genetic and SVD Method for the Prediction of Coastal Wave Impacts

Abstract

Genetic Algorithm (GA) and Singular Value Decomposition (SVD) are deployed respectively for optimal design of Gaussian membership functions of antecedents, and linear coefficients vector of consequents, in ANFIS (Adaptive Neuro-Fuzzy Inference Systems) networks used for prediction of the 3rd month ahead coastal wave impacts. For this purpose time series parameters chosen by GMDH (Group Method of Data Handling) modeling are utilized. The aim of such modeling is to demonstrate how ANFIS network can be coupled with GMDH network and to show how such combination by hybrid GA/SVD designed ANFIS networks result in precise models.