Enhanced Fuzzy Radial Basis Neural Network With Genetic Process Using Fuzzy Inferences

Abstract

Fuzzy Inference-Based Hybridization of a Genetic Algorithm to Enhance the Performance of an Enhanced Fuzzy Radial Basis Neural Network for Better Pattern Recognition and Prediction This proposed framework is comprises fuzzy logic in the body of a regular RBFNN-structured frame that has been optimized through a genetic algorithm. Experimental results shows that the Machine Learning Repository Air Quality data sets offered by the UCI with regular RBFNN, SVM, and Multilayer Perceptron as reference points in comparison with the EFRBNN models. The results are such that EFRBNN outperforms the mentioned methods with 12% accuracy, 9% improvement in F1-score, and 15% mean squared error compared to the standard RBFNN. The genetic process enhanced the optimization of parameters to attain a 20% faster convergence rate. Therefore, it has improved the abilities of this model to confront uncertainties better, and that in the context of noise involved in data, its strengths are 17% robust than these ones in SVM and MLP techniques. These results affirm the fact that EFRBNN may be utilized to manage complex real-time applications based on uncertain data in an imprecise setting.