Personalized content recommendation method, device and equipment

Abstract

The invention relates to a personalized content recommendation method, device and equipment. The personalized content recommendation method is applied to a server, and comprises the following steps: obtaining a plurality of candidate contents corresponding to a target content after receiving a content request of a client for requesting the target content; determining a target model from the standby model and the current reinforcement learning model according to a preset calling rule; wherein the standby model is an algorithm which is used for determining personalized contents based on a plurality of candidate contents and is in an online use state when the current reinforcement learning model is online; the preset calling rule is used for ensuring that a current reinforcement learning model is selected as a first probability of the target model within a first preset time period and is smaller than a second probability of selecting a standby model; determining the target content from the plurality of candidate contents by using the target model, and recommending the target content to the client. According to the scheme, the accuracy of the target content recommended to the user within a certain time period when the reinforcement learning model is online can be improved.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular, to a method, apparatus and device for recommending personalized content. Background technique [0002] In order to improve user experience, various applications, such as music applications, shopping applications, news applications, and video applications, will recommend personalized content that meets the user's preferences for different users. In the related art, after receiving the content request sent by the client for obtaining the target content, the server of various applications obtains multiple candidate contents corresponding to the target content, and then uses a reinforcement learning model to determine the multiple candidate contents. Target content, where the target content is the content to be recommended to the user, so that the determined target content is recommended to the client. [0003] The above reinforcement learning model can be obtai...

AI Technical Summary

Problems solved by technology

[0004] However, during online training, it still takes a certain amount of time for the reinforcement learning model to be trained to accurately determine the target content. In this way, the accuracy of the target content recommended to the user is low within a certain period of time after starting the training.

Images