Novel exploration of machine learning solutions with supervised neural structures for nonlinear cholera epidemic probabilistic model with quarantined impact

Abstract

Cholera is mainly spread by the ingestion of contaminated food or water, especially in areas where poor sanitation is prevalent. The bacteria responsible for cholera, Vibrio cholerae, are observed to multiply in environments lacking proper water treatment and sewage management systems. A novel exploration of machine learning solutions is presented in this paper, with supervised neural structures being applied to a nonlinear stochastic cholera epidemic (SCE) model that incorporates quarantined impact and Brownian motion uncertainty. Artificial neural networks optimized by the Levenberg–Marquardt algorithm (ANNs-LMA) are utilized to predict the dynamics of the SCE model. The probabilistic dynamics of the representative nonlinear SCE model are described in terms of susceptible, infected, quarantined, and recovered individuals, along with the bacterial population represented by the concentration of cholera bacteria in water and food sources. Synthetic data for the execution of ANNs-LMA are generated using the Euler–Maruyama numerical method, with variations in key parameters, including the migration rate into the susceptible group, the transmission rate of cholera through contaminated food and water, the rate at which immunity is lost, natural death rates, the disease progression, and mortality rates among infected individuals, and the recovery or severe disease progression rates among quarantined individuals. The effectiveness of the proposed ANNs-LMA approach is demonstrated by its close alignment with the reference numerical results of the SCE model, as indicated by an error value approaching zero, and is further validated through various assessment metrics, including mean square error-based convergence, adaptive governing parameters, error histograms, and autocorrelation analyses.