Stochastic supervised neuro-architecture design for analyzing vector-borne plant virus epidemics with latency and incubation effects

Abstract

Comprehending the latent and incubation phases is crucial for the proliferation of infectious viruses and serves as a foundation for developing measures to prevent and govern outbreaks. This communication exploits the novel application of feedforward neural network optimizing with the Levenberg–Marquardt scheme (NN-LMQS) for analyzing the dynamics of viral transmission in plants via vectors, with the impacts of the latent and incubation periods. The governing framework consists of a system of delayed differential equations that incorporates both vulnerable and diseased plants, as well as vulnerable and diseased whitefly populations. The reference data set for the vector-borne plant virus epidemic model (VBPV-EM) is derived using the Adams numerical method by adapting variations in the disease propagation rate among infected vectors and susceptible plants, the natural mortality rate of plants, the cumulative growth rates of the vector population due to birth or migration, the vector mortality rate, net plant growth, and time delays. The effective correlations between the designed NN-LMQS and the reference solutions for the nonlinear VBPV-EM substantiate the robustness, proficiency, reliability, and precision of the approach, as endorsed by comprehensive simulation-based assessments, including regression metrics, error histograms, autocorrelation studies, and mean square errors.