一种用户上下文信息驱动的音乐推荐算法  

Music recommendation driven by user context

作　　者：韩丽 杨青[1] 陈奕衡 张敬伟[2] HAN Li;YANG Qing;CHEN Yiheng;ZHANG Jingwei(Guangxi Key Laboratory of Automatic Measure Technology and Instrument,Guilin University of Electronic Technology,Guilin 541004,China;Guangxi Key Laboratory of Trusted Software,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学广西自动检测技术与仪器重点实验室,广西桂林541004 [2]桂林电子科技大学广西可信软件重点实验室,广西桂林541004

出　　处：《桂林电子科技大学学报》2019年第6期460-465,共6页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(61462017,61862013,61662015);广西自然科学基金(2017GXNSFAA198035)。

摘　　要：针对用户的个性化音乐欣赏需求,兼顾用户所处的特定场景,提出了一种集成协同过滤与隐马尔可夫模型的音乐推荐算法(CF-HMM)。以用户行为为参考维度,基于用户历史行为,优化用户相似度及其计算;综合考虑用户上下文信息和历史行为建立隐马尔可夫模型,并获取最大音乐类别概率序列;融合最大音乐类别概率序列与相似用户偏好,实现目标用户的多样化音乐推荐。真实数据集的实验结果表明,提出的推荐算法是有效的。Considering users’personalized requirements on recommendation,a music recommendation method under users’specific context is proposed,which integrates both collaborative filtering and the hidden markov model for better performance.First,an optimized metrics for computing user similarity by making a comprehensive consideration between users’historical behavior and music category is provide.Second,both user context and users’historical behavior are exploited to establish a hidden markov model for obtaining a music category sequence with the maximum observation probability.Third,both the output category sequence and the preferences of those similar users are integrated to discover the recommended items with diversification.Finally,we designed experiments to verify the effectiveness of the proposed method.

关 键 词：用户上下文 音乐类别 隐马尔科夫模型 概率序列 多样化推荐 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]