A comparison of BA, GA and PSO in estimation of the Effect of Higher-Order Aberrations on Visual Function in Keratoconic Eyes using a FCM-ANFIS approach

Optimizing ANFIS model is a complex task of greatimportance in the hybrid neuro-fuzzy system field of research. Inthis work we tackle this problem with three optimizationalgorithms; Bat Algorithm (BA), Genetic Algorithm (GA), andParticle Swarm Optimization (PSO). We intend to show thesuperiority (time performance and quality of solution) of the newmetaheuristic BA over other more standard optimizationalgorithms. Therefore, the present research adopts each of thesethree metaheuristic techniques to obtain appropriate parametersettings for Fuzzy C-Means clustering (FCM) and integrates theAdaptive Network-based Fuzzy Inference System (ANFIS) forestimating the visual function of Keratoconic eyes. Ourconclusions clearly establish the advantages of the newmetaheuristic BA over the other algorithms in the optimizing ofFCM-ANFIS model for this clinical problem. The results showthat in the case of 50 bats and 100 iterations, the performanceindices of BA; RMSE, MAPE and Correlation were calculated as0.092637, 0.013722 and 0.868844 respectively.

Keywords:

Adaptive network-based fuzzy inference system ANFIS) , Fuzzy C-Means Clustering , Bat algorithm optimization Genetic Algorithm , Particle Swarm Optimization , Keratoconus