A machine learning radial basis deep neural network procedure for the solutions of the nonlinear breast cancer model

Abstract

This study provides the numerical performances of the nonlinear mathematical breast cancer (BC) system by designing a novel computational radial basis deep neural network structure. A process of two hidden layers with 20 and 42 numbers of neurons is used with the radial basis activation function for solving the model. The mathematical BC model is categorized into different cells of cancer stem, healthy, tumor, immune, and excess estrogen. The optimization is performed through the Bayesian regularization for solving the BC model. A dataset is constructed by applying the explicit Runge-Kutta to lessen the mean square error by using the data performance, which is divided into training 78%, testing 12%, and validation 10%. The stochastic neural network process is based on two hidden layers, radial basis activation function, 20 and 42 numbers of neurons, and feed forward neural network for solving the BC model. The scheme's correctness is observed through the overlapping of the results and negligible absolute error values around 10(-05) to 10(-07) along with different statistical performances.