首页> 外文期刊>電子情報通信学会技術研究報告 >Utilizing Users' Watching Sequences and TV-programs' Metadata for Personalized TV-program Recommendation
【24h】

Utilizing Users' Watching Sequences and TV-programs' Metadata for Personalized TV-program Recommendation

机译:利用用户的观看顺序和电视节目的元数据进行个性化的电视节目推荐

获取原文
获取原文并翻译 | 示例
       

摘要

Recently, the explosive growth of digital video contents including IPTV (Internet Protocol Television) has led to the need of recommendation system to guide users among various and huge amount of entertainment movies, live-TV or related services that are called TV programs in general. Consequently, recommendation system has become a general tool to support user's decision in making choice. Most of the ever-proposed algorithms focus on the prediction accuracy; however, we also have to support the diversity of the recommendation results to surprise users in order to widen their choices that might be just missed if the accuracy is only focused on. In this paper, we introduce a new model-based top-K recommendation algorithm called "watch-flow algorithm" for selecting the next K highest potential TV programs that user might like. Our model utilizes users' watcning sequences and TV program metadata to identify the recommending value for each TV program. Furthermore, this model is also capable of giving a personalized recommendation for a specific user based on his/her watching sequence, as well as capable to improve the prediction accuracy and the diversity. We apply our algorithm on a random sample of users' watching sequences in a dataset collected from real users' log. According to the experimental results, our proposed method shows better performance in recommendation than that of ever-proposed algorithms in terms of higher accuracy while keeping the coverage of programs in high rate.
机译:近来,包括IPTV(互联网协议电视)在内的数字视频内容的爆炸性增长,导致需要推荐系统来指导用户在各种娱乐电影,实况电视或通常称为电视节目的相关服务中进行指导。因此,推荐系统已成为支持用户决策的通用工具。大部分提出的算法都专注于预测精度。但是,我们还必须支持推荐结果的多样性,以使用户感到惊讶,以扩大他们的选择范围(如果仅关注准确性,可能会错过这些选择范围)。在本文中,我们介绍了一种新的基于模型的top-K推荐算法,称为“观看流算法”,用于选择用户可能喜欢的下K个潜力最大的电视节目。我们的模型利用用户的等待顺序和电视节目元数据来识别每个电视节目的推荐值。此外,该模型还能够基于特定用户的观看顺序来为其提供个性化推荐,并且能够提高预测准确性和多样性。我们将算法应用于从真实用户日志收集的数据集中的用户观看序列的随机样本。根据实验结果,我们提出的方法在保持较高的覆盖率的同时,在准确性方面表现出比以往提出的算法更好的推荐性能。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号