The outbreak of new coronavirus pneumonia in 2019, which is called Corona Virus Disease 2019 (COVID‐19), has been the most serious infectious disease pandemic in the world in 100 years. It is a major public health emergency with the fastest spreading speed, the widest infection range and the most difficult to prevent and control. Currently, the spread of COVID‐19 is still global, and the epidemic situation in some countries is still very serious. Many scholars have also started to crawl, summarize, analyze various kinds of historical data emerging on the network, and use various algorithms to design the epidemic prediction model of new coronary pneumonia. In this paper, based on the confirmed, dead and cured cases of COVID‐19 in the United States obtained from the Center for Systems Science and Engineering of Johns Hopkins University, SIR models, logistic regression as well as support vector regression algorithms in machine learning are used to simulate and predict the development of the epidemic, and the accuracy of each prediction model is compared. In order to provide more accurate reference for the follow‐up epidemic warning and prevention and control.