大屏端智能实时推荐系统建设实践

Construction Practice of Large-screen Terminal Intelligent Real-time Recommendation System

本文主要解决如何在海量的视频内容中快速为用户过滤出感兴趣内容的问题,如何在大数据计算平台的有力支撑下进行算法模型的选择和训练,如何使用算法评估、A/B测试等方法不断优化模型以及时捕捉用户兴趣的变迁,重点研究了用户、视频和场景信息的关系,并根据大屏终端实际用户是多个家庭成员的特点,开发了针对家庭用户的分时分群实时推荐算法,可为用户提供精准推荐内容,有效提升了用户观看体验。

This article mainly solves the problem of quickly filtering out content of interest for users in massive video content,how to select and train algorithm models under the strong support of big data computing platform,and how to use algorithm evaluation,A/B testing,and other methods to continuously optimize the model to timely capture the changes of user interests,and focuses on the relationship between users,videos and scene information.Considering the characteristics of multiple family members being the actual users of the large-screen terminal,it has developed a real-time recommendation algorithm for family users based on time sharing and group segmentation.This provides users with precise recommendation content and effectively enhances the viewing experience.