Behavior prediction model training method and device

Abstract

The embodiment of the specification provides a behavior prediction model training method, comprising: for a target object, obtaining a plurality of positive samples formed based on a plurality of seedusers, any first positive sample comprising a first user feature and a positive example label corresponding to a first seed user, and a second positive sample comprising a second user feature and a negative example label corresponding to a second seed user, wherein the label indicates that the corresponding user is a user determined to make a specific behavior on the target object; determining anoutlier score of the first seed user by adopting an unsupervised outlier detection algorithm based on respective user characteristics of the plurality of seed users, and taking the outlier score as afirst training weight for the behavior prediction task; and performing first training on a first behavior prediction model by utilizing a plurality of positive samples and a plurality of pre-acquirednegative samples, specifically inputting a first user feature into the first behavior prediction model, determining behavior prediction loss by combining an obtained behavior prediction result and the positive example label, and using the first training weight is used to carry out weighting processing on the first behavior prediction model so as to train the first behavior prediction model.

Description

[0001] This application is a divisional application based on the patent application with the title of "Training Method and Device for Behavior Prediction Model" filed on August 14, 2020, with application number: 202010819192.4. technical field [0002] The embodiments of this specification relate to the field of computer technology, and in particular to a method and device for training a behavior prediction model. Background technique [0003] Currently, the service platform usually recommends or pushes business objects such as products or content to users, for example, recommending some online courses, clothing products, advertisement pictures, and the like. With the cumulative growth of the number of business objects and the continuous emergence of new business objects, in order to improve user experience, it is necessary to promptly and accurately recommend business objects that meet their needs and preferences to users. Accordingly, the service platform can use machine le...

Images