Data-driven optimal modeling and prediction of human brucellosis in China

Brucellosis, a typical zoonotic disease, has long been a public health concern in China due to its high incidence and wide range of infections. Quantitative studies of epidemiological data are essential for understanding the dynamics of disease transmission, developing intervention programs, and eradicating infectious diseases. This paper proposes an improved data-driven method called sparsity-promoting Hankel dynamic mode decomposition (HDMDsp). This method is applied to characterize the epidemiological features of human brucellosis in China. By performing dynamic mode decomposition (DMD), this method extracts the spectral characteristics of the underlying system, allowing for the analysis of dynamic changes in infectious diseases. To improve prediction accuracy and simplify computation, this paper introduces a sparsity-promoting technique–elastic network regularization–to select the dominant model for reconstructing the evolutionary system of disease transmission. Additionally, a Hankel matrix is constructed via time-delay embedding to project the disease data into high dimensions, thereby enhancing the applicability of DMD. A study on human brucellosis in China demonstrates that the HDMDsp algorithm is highly effective at describing the spatiotemporal distribution of the disease and accurately predicting its transmission dynamics.