Abstract
Malaria remains one of the leading causes of global morbidity and mortality, with millions of cases and fatalities annually. Effective intervention strategies by public health authorities and medical practitioners necessitate a robust understanding of disease transmission dynamics. This study presents a novel framework for modeling malaria transmission dynamics by integrating temperature and altitude-dependent transmission functions into a compartmental SIR-SI model. A key innovation lies in the introduction of a new transmission function that explicitly captures environmental dependencies, enhancing realism in the modeling of disease spread. We conduct steady-state analysis of the system, establishing the stability criteria for both disease-free and endemic equilibria through linearization techniques. We used a novel transmission function to model the dependence on temperature and altitude. To address the challenge of accurate parameter estimation, we develop a comparative learning framework using ANNs, RNNs, and PINNs, with PINNs standing out by embedding epidemiological dynamics into the training process. This enables physics-constrained parameter inference, significantly enhancing predictive performance over purely data-driven approaches. Additionally, we implement Dynamic Mode Decomposition (DMD) to derive a data-driven transmission risk index from infection trajectory data, providing a novel and interpretable metric for real-time risk assessment.