A Numerical Treatment Through Bayesian Regularization Neural Network for the Chickenpox Disease Model

Abstract

Objectives: The current research investigations designates the numerical solutions of the chickenpox disease model by applying a proficient optimization framework based on the artificial neural network. The mathematical form of the chickenpox disease model is divided into different categories of individuals, susceptible, vaccinated, infected, exposed, recovered, and infected with/without complications. Method: The construction of neural network is performed by using the single hidden layer and the optimization of Bayesian regularization. A dataset is assembled using the explicit Runge-Kutta technique for reducing the mean square error using the training 76 %, while 12 %, 12 % for validation and testing. The whole stochastic procedure is based on logistic sigmoid fitness function, single hidden layer structure with thirty neurons, along with the optimization capability of Bayesian regularization. Finding: The designed procedure's correctness and reliability is observed by results matching, negligible absolute error around 10−04 to 10−06, regression, error histogram, and state transmission. Moreover, the best performance values based on the mean square error are performed as 10−09 to 10−11. Novelty: The current neural network framework using the construction of a single hidden layer and the optimization of Bayesian regularization is applied first time to solve the chickenpox disease model. © 2025 Elsevier Ltd