Accompanied by the rapid economic and social development, there is a phenomenon of the crazy spread of many infectious diseases. It has brought the rapid growth of the number of people infected with hand-foot-and-mouth disease (HFMD), and children, especially infants and young children's health is at great risk. So it is very important to predict the number of HFMD infections and realize the regional early-warning of HFMD based on big data. However, in the current field of infectious diseases, the research on the prevalence of HFMD mainly predicts the number of future cases based on the number of historical cases in various places, and the influence of many related factors that affect the prevalence of HFMD is ignored. The current early-warning research of HFMD mainly uses direct case report, which uses statistical methods in time and space to have early-warnings of outbreaks separately. It leads to a high error rate and low confidence in the early-warning results. This paper uses machine learning methods to establish a HFMD epidemic prediction model and explore constructing a variety of early-warning models. By comparison of experimental results, we finally verify that the HFMD prediction algorithm proposed in this paper has higher accuracy. At the same time, the early-warning algorithm based on the comparison of threshold has good results.
Keywords: early-warning model; genetic algorithm; hand-foot-and-mouth disease; neural network.