直播电视推荐系统的评分预测算法研究

Research on Rating Prediction Algorithm of Live TV Recommender Systems

伴随电视频道的不断增加,推荐系统在直播电视领域应用成为研究热点.然而,直播电视独特的播放和收视方式使得传统的VOD(Video On Demand)推荐系统无法直接应用,已有的推荐频道的方法不关注正在播出的节目状态从而影响了推荐准确率,而推荐节目的方法难以应对节目冷启动.为此,本文提出了一种融合频道推荐和节目推荐的评分预测算法OFAP(Over the First by Adding Preference).首先,利用聚类方法对每个用户实现差异性的收视时段划分,构建他们的频道-时段偏好矩阵和预推荐评分权重矩阵;其次,提出一个评分替代策略使得已有的推荐节目的算法能够应对节目冷启动,从而实现预推荐;最后,通过融合用户偏好、预推荐评分权重与预推荐结果,构建评分预测函数,将预推荐算法的评分预测结果作为评分预测函数的训练样本.实验表明,采用Precision@N和Recall@N作为评价标准,本文所提方法OFAP明显优于对比算法.

With the increase of TV channels,the application of recommender systems in the field of live TV has become a research hotspot.However,a traditional VOD(Video on Demand)recommender system is unable to be directly applied in live TV because of its special way of broadcasting and watching,and the existing methods of recommending channels do not pay attention to status of TV shows being broadcasted,which affects recommendation accuracy,and the methods of recommending programs are difficult to handle cold start of TV shows.Therefore,this paper proposes a rating prediction algorithm by fusing TV channel recommendation method and TV program recommendation method OFAP(Over the First by Adding Preference).Firstly,we construct different channel-time preference matrix and rating weight matrix of pre-recommendation for each user by clustering their viewing logs.Secondly,we propose a rating strategy to alleviate the cold-start problem of TV programs for existing program recommendation algorithms,and we adopt one of them to perform pre-recommendation.Finally,we combine user’s preference,rating weight and rating of pre-recommendation to construct a prediction function,which is trained with the results of pre-recommendation.Experiments on industrial datasets show that the proposed model OFAP significantly outperforms baseline algorithms when Precision@N and Recall@N are adopted as criterias.