This paper proposes a method utilizing an evolutionary algorithm (EA) to improve the learning process within the artificial neural network (ANN). The study utilized two models: the neuroevolutionary algorithm and the GANet tool. These algorithms were trained on two clinical diabetes datasets to discern health conditions. Each Feedforward Artificial Neural Network (FANN) layer in a neuroevolutionary algorithm is broken into multiple ANNs according to the number of neurons. The evolutionary technique later utilizes crossover and mutation operations to produce an ideal FANN. Enhanced artificial neural networks were subsequently incorporated to provide an innovative and robust functional artificial neural network for practical use. The GANet tool offers a novel mutation methodology. This method enables neurons to transform into multi-layered subnetworks. The tests utilizing diabetes data attained enhanced prediction accuracy. This demonstrates the importance of optimizing neural networks. It creates a strong basis for the automation of AI design and the advancement of medical diagnostics in the future. The proposed tools exhibited superior efficiency relative to alternative methods such as decision trees and support vector machines.
Stepanyan et al. (Tue,) studied this question.