A radial basis Bayesian regularization procedure for the Lassa virus mathematical model

Abstract

The purpose of current research investigations is to perform the numerical investigations of the Lassa virus model by using the computing stochastic paradigms. The Lassa virus was identified first in Nigeria, and 20 years later, the mathematical Lassa virus model has been developed by including different factors such as population of human to human, rodent to human, and environmental influences. A single hidden layer neural network structure using a radial basis function, fifteen neurons, and optimization with Bayesian regularization is presented to solve the Lassa virus mathematical model. The construction of the dataset is performed by the explicit Runge–Kutta scheme, which reduces mean square error by dividing the statistics into training as 74%, while 14% for authentication and 12% for testing. The correctness of proposed scheme is authenticated by solving three different model cases including comparison of the results that are 6–8 decimal places, best training performances around 10⁻¹¹–10⁻¹⁴, and absolute errors found as 10⁻⁰⁶–10⁻⁰⁸. The reliability of the designed stochastic neural network is obtained by using different tests including correlation, state transitions, and error histograms.