Media resource recommendation method, and multi-target fusion model training method and device

Abstract

The invention provides a media resource recommendation method and a multi-target fusion model training method and device, belongs to the technical field of machine learning, and can be applied to various scenes such as cloud technology, artificial intelligence and intelligent traffic. The method comprises the steps of determining business target estimation information based on object features of a target object, resource features of media resources and associated features; determining object type information based on the object features; determining recommendation information of the media resources based on the business target estimation information and the object type information; and recommending the media resource to the target object in response to the condition that the recommendation information meets the recommendation condition. According to the technical scheme, the prediction effects of the business targets can be fused on the basis of the probability that the objects belong to the object types, so that the fusion modes of the prediction effects of the objects and the business targets of different object types are different, the accuracy of the fusion result is improved, and the problem of'tradeoff 'can be effectively solved.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular to a media resource recommendation method, a multi-object fusion model training method and a device. Background technique [0002] The multi-objective ranking model is widely used in recommendation systems. The multi-objective ranking model obtains additional benefits that the single-objective model cannot obtain by simultaneously optimizing the effects of multiple business objectives. The effects of business objectives include click-through rate, conversion rate and Click conversion rate etc. However, the problem often encountered in multi-objective modeling is that the effect of some business objectives has been improved, but the effect of other business objectives has become worse, that is, it is difficult to make the effects of all business objectives better at the same time. The problem is simply referred to as the "one ebb and flow" problem. [0003] A...

AI Technical Summary

Problems solved by technology

[0004] However, although the above scheme can make each business goal fit better, it is necessary to integrate the estimated effects of different business goals at the stage of model prediction recommendation probability. The method is not good, resulting in low accuracy of fusion results, and the problem of "one ebb and flow" occurs

Images