A novel deep learning technique for solving dynamical model of infertile women with Polycystic Ovarian Syndrome Using the combination of radial basis and log-sigmoid functions

Abstract

The focus of this study is to introduce the novel process of deep neural network (DNN) for the numerical solution of dynamical model based on infertile women with polycystic ovarian syndrome (PCOS). The model consists of five classes such as, Susceptible Women (S), Infertile Women (I), Women under Treatment 1 ( Women with Treatment 2 (and Recovered Women (R). The dual hidden layer neural network procedure is applied using the combination of radial basis function and sigmoid function for the numerical approximations of the nonlinear model. The twenty neurons for radial basis function and forty neurons for sigmoid function are used. The acquired numerical results are validated through the comparison of the solutions. The optimization is implemented by employing a competent scale conjugate gradient procedure. The verification of the proposed numerical method is observed through the reference results, and the reduced absolute error around is obtained. The reference database is achieved using the Runge-Kutta method of order 4. The method is proved reliable when trained, tested and validated using the reference dataset that consists of proportions 72%, 14% and 14%. After using the DNN approach, results are validated and displayed through MSE, regression, fitness, error and comparison plots.