Employing Levenberg-Marquardt backpropagation(LMB) neural network, the system of three species nonlinear equations are illuminated by designing an integrated numerical computing-based plot. The proposed dynamical system comprises of two competing species which are growing logistically in nature and, the third species is predating with Holling type II functional response mode on second species and also acts host for the first prey species. Besides, the prey species protect themselves to refuge high predation. The designed LMB neural network has been used to exhibits the solutions of the dynamical frame work. In each case of the species, a reference dataset of the planned LMB neural network is initiated in comparison of Adam numerical program. The approximate results of the food web system are displayed within the training, confirmation and testing strategies to redesign the neural network to minimize the mean square error (MSE) function employing the designed LMB. The investigations depend on the corresponding achievements and the examinations based on MSE out comes, correlation, regression and error histograms signify the proficiency, rightness as well as the potency of the suggested LMB neural network conspire.