Learning anfis parameters using elephant herding optimization algorithm

Fuzzy systems are an artificial intelligence and control approach based on the principles of fuzzy logic. Unlike traditional logic systems, they operate using fuzzy sets, fuzzy rules, and fuzzy inference mechanisms instead of linear and exact rules. They provide a more flexible approach for modeling and controlling complex and uncertain systems. Fuzzy systems are inherently expert-based, and it may not always be possible to reach an expert when adjusting their parameters, or parameters may not be determined with sufficient accuracy in the face of large data sets. For this reason, adjusting the relevant parameters using machine learning or optimization algorithms has become a popular research topic today. For this purpose, the Adaptive Network-Based Fuzzy Inference System (ANFIS), which learns from a training data set with input and output, is frequently used in the literature. In this thesis study, the Elephant Swarm Optimization (EHO) algorithm is proposed for learning ANFIS parameters and is named ANFIS-EHO. EHO is a meta-heuristic optimization algorithm that mimics elephant herd behavior, where the lead elephant advances towards the correct solution and the follower elephants can perform a global search by following it. The success of the proposed method is measured on four nonlinear static system identification problems. As an experimental study, the parameters of the ANFIS system with 4-9-16 fuzzy rules were learned using the EHO algorithm, and this study was repeated 30 times. RMSE error values and R2 values are reported. The best results were obtained for the 16-rule structure for all problems. To discuss the success of the ANFIS-EHO algorithm, the parameters of the 16-rule ANFIS structure were learned using the Genetic Algorithms and Particle Swarm Optimization algorithms. For each nonlinear static system, the algorithms were run 30 times and the results were statistically compared using the Wilcoxon signed-rank test. According to the results, the ANFIS-EHO algorithm was found to produce more successful and consistent results.