A radial basis neural network using the optimization of Bayesian regularization of the malaria disease model

Abstract

The current research investigations are used to provide a stochastic computational radial basis function neural network (RBFNN), which is one of the kinds of the artificial neural network that applies radial basis as an activation function. The optimization is performed through the Bayesian regularization (BR) and the proposed solver is named as RBFNN-BR in order to solve the malaria disease model (MDM). The mathematical form of the MDM is categorized into host and vector populations that are based on pesticides and medication. A dataset is constructed based on the explicit Runge-Kutta scheme, which is used to reduce the mean square error (MSE) by selecting the data for testing 0.15, validation 0.12 and training 0.72 for the numerical solutions of the mathematical MDM. The solution of the MDM is presented by taking twenty numbers of neurons, RBF as an activation function in the hidden layers, RB for the optimization, and data selection based different values. The correctness of the RBFNN solver is observed by using the comparison with the published literature results accurateness and the reference solutions for solving the MDM. Moreover, the negligible absolute error performances also approve the precision of the scheme. The competency of the proposed solver is authenticated by using different performances in the sense of MSE, regression and error histogram.