Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
Table 2 The average AUC score. This is accurate enough to find the principle of our SEIR model, the ranking model of all data, on the condition that we eliminate the people with state I that do not have any surrounding features measured in up to three layers. The prediction divided people into two groups by probability of 0.5: either they will or will not be infected. The result of feature importance shows the contact time to the high-risk location which is more relevant for judging whether someone has been infected or not, and where \(X_I\) and \(X_E\) are the total numbers of state I and E in the first three surrounding layers, respectively.