Stochastic Performances of the Mathematical System Based on the Prevalence Prediction of Allergies

The objective of this study is to conduct numerical simulations of a six-compartment model in order to predict the likelihood of developing allergies. The research categorizes the division of model into four classes of human population, along with two allergen groups, namely inhalants and food. The study aims to develop a predictive model for allergen sensitization occurrence using a system of nonlinear differential equations. The numerical presentation of the system's solution is achieved through the utilization of the stochastic computational artificial neural network (ANN) and the Levenberg–Marquardt backpropagation (LMBP). The neural network process involves presenting three types of statistics: training, testing, and sampling. The allergy occurrence prediction model has a preferred statistical accuracy of 72% for training data and 14% for both authentication and testing data. The accuracy of the stochastic scheme is evaluated by comparing its performance with the proposed and reference solutions. A small absolute error is an indication of the scheme's correctness. The research observes the reliability of the stochastic solver by utilizing various statistical operators.