Method and device for carrying out crowd division on users and method and device for training multi-task model

Abstract

The embodiment of the invention provides a method and device for carrying out crowd division on users and a method and device for training a multi-task model, and the method comprises the steps of obtaining a user total set, and enabling possible values of a plurality of feature variables of the users to form a feature value set; constructing a relational network diagram, wherein a single node corresponds to one feature value in the feature value set, wherein a connection edge connecting the two nodes has an edge attribute value, and the edge attribute value is determined according to the number of users in a user total set and having two feature values respectively represented by the two nodes at the same time; obtaining node embedding vectors respectively corresponding to each node in the relational network graph through a graph embedding mode; selecting a plurality of feature value subsets from the feature value set according to the similarity between the node embedding vectors; and taking the plurality of feature value subsets as screening conditions of the user total set, and selecting a plurality of user subsets from the user total set. A good distinguishing effect of different crowds can be realized, and the prediction accuracy of the model is improved.

Description

technical field [0001] One or more embodiments of this specification relate to the field of computers, and in particular to methods and devices for classifying users into groups and training multi-task models. Background technique [0002] Currently, in order to improve business goals, it often involves the prediction of the user's business goals, for example, when recommending items to the user, the prediction of the user's click rate on the recommended items. In the prior art, when performing the above prediction, users are not divided into groups, and different preferences of different groups of people cannot be reflected in the prediction, so the prediction result is not accurate enough. [0003] Therefore, it is necessary to provide a method for classifying users into groups so as to achieve a good effect of distinguishing different groups of people; and to provide a method for model training based on the divided groups of people so as to improve the prediction accuracy...

AI Technical Summary

Problems solved by technology

In the prior art, when the above prediction is made, the users are not divided into groups, and the different preferences of different groups of people cannot be reflected in the prediction, so the prediction results are not accurate enough

Images