基于DNN算法的移动视频推荐策略  

DNN Based Mobile Online Video Recommendation

作　　者：陈亮[1] 汪景福[1] 王娜[1] 李霞[1] 

机构地区：[1]深圳大学信息工程学院,广东深圳518060

出　　处：《计算机学报》2016年第8期1626-1638,共13页Chinese Journal of Computers

基　　金：国家自然科学基金(61502315;61309030);广东省自然科学基金(2015A030310366);深圳大学科研启动项目(201558)资助

摘　　要：该文针对移动互联网在线视频服务个性化视频推荐问题,提出了一种基于深度学习(Deep learning)模型的内容推荐策略.深度学习在图像、语音和自然语言处理等领域获得的突破性进展,为在线视频服务的推荐策略研究提供了基础.该文工作通过在传统基于内容推荐的基础上,引入深度神经网络(Deep Neural Network,DNN)词向量方法,根据视频的媒资信息和用户的历史行为,将视频特征和用户特征映射在高维向量空间.在构建用户正负行为与视频向量的余弦距离模型基础上,筛选过滤对所推荐视频感兴趣的用户群,并通过移动互联网应用的消息推送功能提示移动终端用户观看所推荐的内容.基于在大规模移动视频服务系统中的离线和在线实验,该文所提出的基于DNN算法的推荐策略,相比随机方法、ContentKNN以及ItemCF等算法,在点击率方面平均分别获得106%、41%和57%的相对提升,在覆盖率方面一定程度上避免了推送活跃用户的偏颇问题,从整体上得到了较好的推荐效果.We study the personalized video recommendation strategy in mobile online video services by introducing the deep learning method in the content-based recommendation.Deep learning algorithms have made a good progress in image recognition,speech recognition and natural language processing,which lay the foundation of the recommendation study in online video services.By mapping each video（according to its description information）and each user（according to her/his history behavior）into a high dimensional vector space with word embedding methods,we can filter the target user shaving potential interest in candidate videos based on the vector relationship model,and push notifications of recommended videos to them accordingly.With offline and online experiments in a large-scale mobile online video system,we evaluate our proposal and compare it with the random method,contentKNN algorithm and ItemCF algorithm.The experiment results show that our proposal＇s performance has been improved by as much as106%,41% and 57%respectively on the click-through rate in the mobile online video services.

关 键 词：移动互联网 深度学习 词向量 消息推送 基于内容推荐 机器学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]