User Tag Determination Method and Device

Abstract

This application relates to the field of data mining technology and provides a method and apparatus for determining user labels. The method first invokes at least two binary classification models to perform probability prediction processing on multiple structured features of a target user, determining the first probability values ​​of at least two attribute values ​​of the target label output by each binary classification model. Then, based on the multiple structured features of the target user, it determines at least one preset prior rule triggered by the target user. Next, based on each triggered preset prior rule, it determines the second probability value of each attribute value. Finally, by combining the first and second probability values ​​of each attribute value, it determines the attribute value of the target label. Therefore, by invoking binary classification models and combining preset prior rules to process the structured features of the target user, it can automatically predict the missing target labels of the target user. This application can automatically perform batch labeling of a large number of users, improving the efficiency of user label determination.

Images