Heterogeneous Data Mining for Planning Active Surveillance of Malaria

Malaria is one of the most serious diseases in the world, which is densely distributed in poverty and remote areas. In the prevention and control of malaria, active surveillance is more efficient than passive surveillance to discover the incidences timely and accurately. However, it is always faced with the challenge of how to allocate the limited sources, such as medical staff and medicine, appropriately so as to achieve a maximum infect. In this paper, we propose a novel method to characterize the spatiotemporal patterns of infection risk for active surveillance planning. Specifically, we propose a temporal heterogeneous diffusion network model to discover high risk areas timely, and a mixture optimization method to find high risk areas accurately. The validation on existing real-world data shows that our method outperforms the existing state-of-the-art both in terms of infection risk prediction and planning of active surveillance under different thresholds.