FRACTIONAL-ORDER PREY-PREDATOR MODELS WITH PARAMETER ESTIMATION VIA FRACTIONAL PHYSICS-INFORMED NEURAL NETWORKS (FPINNS)

This article examines the fractional prey-predator model through the use of fractional physics-informed neural networks (fPINNs). The methodology focuses on the implementation of fPINN algorithms to approximate solutions to systems governed by fractional-order differential equations. Furthermore, key parameters of the models, based on Holling's-type functional responses I, II, and III, were estimated with fPINNs. A detailed study is conducted on the impact of data noise, various data sets, and if only one population is available (prey), then how to predict the other population (predator). Our computational experiments indicate that fPINNs are extremely effective, both in parameter estimation and in accurately forecasting prey-predator populations.