Mathematical modeling of chikungunya transmission dynamics and control using disease-driven machine learning

Abstract The Chikungunya virus (CHIKV) is a significant vector-borne threat; however, few models successfully integrate rigorous theoretical guarantees with efficient, reliable predictive tools. This study aims to address this gap by introducing a nonlinear seven-compartment human-mosquito model. This model includes susceptible, exposed, infected, and recovered humans, as well as susceptible, exposed, and infected mosquitoes. It accounts for latency stages on both sides of the transmission cycle, an epidemiological aspect that is often overlooked in existing CHIKV models. The mathematical well-posedness of the model is established by demonstrating the positivity and boundedness of solutions and calculating the basic reproduction number, denoted as $$\mathcal {R}_0$$ . Furthermore, it is shown that the disease-free equilibrium is locally asymptotically stable for $$\mathcal {R}_0 < 1$$ and, importantly, globally asymptotically stable across the feasible region when $$\mathcal {R}_0 \le 1$$ . This finding represents a stronger result than the local stability outcomes observed in similar models. The main methodological contribution is the development of a disease-informed machine learning framework based on a random projection neural network (RPNN) as a computationally efficient alternative to conventional deep neural networks (DNNs) to solve the CHIKV system. This is compared against a DNN benchmarked head-to-head on different epidemiological scenarios, with the RPNN consistently achieving higher accuracy, stability, and computational cost, thus offering the ability to serve as a surrogate solver for epidemic models more generally. We also validate the framework using real chikungunya surveillance data from Colombia, demonstrating that the fitted model reproduces the observed trajectory of the outbreak. The parametric analysis shows that higher transmission rates accelerate the onset of the disease, lower recovery rates prolong and enhance the infections, and a higher mortality rate of mosquitoes is an effective lever for outbreak control. The RPNN-based prediction framework and real data validation provide a joint and computationally efficient tool for analysis and control of chikungunya outbreaks as a result of the theoretical guarantees.