Performance Comparison of Nonlinear Kalman Filters in Epidemic Tracking on Networks

Abstract

The problem of tracking epidemic spreading on networks is relevant to the control of morbidity. The transmission dynamics of an epidemic can be described by a simple compartmental model. Specifically, the estimation of epidemic spreading on networks can be achieved by a nonlinear Kalman filter, which is a tool for state estimation of nonlinear systems. In this article, epidemic spreading on networks is described by compartmental models, such as susceptible–infected–susceptible, susceptible–infected–recovered, and susceptible–infected–recovered–susceptible models. The dynamics of epidemic spreading on homogeneous networks, including Erdos–Renyi network, Watts–Strogatz network, and Newman–Watts network, are estimated by several nonlinear Kalman filters, including extended Kalman filter, unscented Kalman filter, and third-degree and fifth-degree cubature Kalman filters. The performance comparison in terms of accuracy and stability forms a guideline of utilizing nonlinear Kalman filters for tracking epidemic spreading. The theoretical analysis has been validated through numerical experiments.