个性化音乐推荐系统的设计与实现
发布时间：2018-11-13 17:06
【摘要】：大数据浪潮下,互联网在方便人们获取数据的同时,也将人们置于海量信息选择的困境中。传统的搜索引擎虽能过滤掉一部分无关信息,但从大量搜索结果中逐一挑选也是一件费时的事情。学术界和工业界就这一问题具有一致观点,即推荐系统是当前缓解大数据灾难最为有效的方式。音乐之于人类而言,已经成为生活的重要部分。随着互联网服务的发展,人们对音乐的消费模式也发生了巨变,以用户为中心的相关技术已成为目前音乐服务的主流技术。社会化标签既反映了资源的特征属性,也体现了用户的兴趣所在。同时,用户在不同时间对标签的感兴趣程度是在不断变化的。本文将结合这两个特性,提出一种基于标签和时间加权的推荐算法模型。利用用户行为日志,建立时间迁移下的用户兴趣模型,并通过社会化标签标识歌曲内容,然后依据用户兴趣模型与备选歌曲的标签特征的匹配程度来生成推荐列表。同时标签可以很好地解释推荐该项目的原因,提高用户的接受度。为了解决数据稀疏性的问题,本文还加强了对隐式数据的利用。为解决新用户冷启动问题,本文提出了一种基于用户基本信息查找相似用户的解决方案。首先利用三部图的物质扩散算法衍生的相似度计算法则,将系统中具有相似行为的用户进行聚类,再将用户的基本信息与其类别建立关系,由此便可根据新用户基本信息,找到其相似用户群,为其赋予初始标签偏好特征,从而为新用户生成个性化推荐列表。最后,本文针对个性化音乐推荐系统的需求设计了推荐系统总体架构,划分了各功能子模块,对各模块进行了详细设计与实现。在交互模块,所设计的音乐可视化界面能直观地展示项目间的关联性,提高用户体验。
[Abstract]:Under the tide of big data, the Internet not only makes it convenient for people to obtain data, but also puts people in the dilemma of mass information choice. Although traditional search engines can filter out some irrelevant information, it is also time-consuming to select one by one from a large number of search results. Academia and industry agree that recommendation system is the most effective way to alleviate big data disaster. Music has become an important part of human life. With the development of Internet service, the consumption mode of music has changed greatly, and the user-centered technology has become the mainstream technology of music service. Social tags not only reflect the characteristics of resources, but also reflect the interests of users. At the same time, the extent to which users are interested in labels is constantly changing at different times. In this paper, a recommendation algorithm model based on label and time weighting is proposed. By using user behavior log, the user interest model under time migration is established, and the content of songs is identified by social tags, and then the recommendation list is generated according to the matching degree between user interest model and label characteristics of alternative songs. At the same time, the label can explain the reason why the project is recommended and improve the acceptance of the user. In order to solve the problem of data sparsity, we also strengthen the use of implicit data. In order to solve the cold start problem of new users, this paper proposes a solution for finding similar users based on user's basic information. Firstly, by using the similarity calculation rule derived from the material diffusion algorithm of the three-part graph, the users with similar behavior in the system are clustered, and then the basic information of the user is related to its category, and then the basic information of the new user can be based on the basic information of the new user. Find the similar user group, assign the initial label preference feature, and generate the personalized recommendation list for the new user. Finally, according to the requirements of the personalized music recommendation system, this paper designs the overall framework of the recommendation system, divides each functional sub-module, and designs and implements each module in detail. In the interactive module, the design of the music visualization interface can show the relationship between projects intuitively and improve the user experience.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2018
【分类号】：TP391.3
本文编号：2329780
[Abstract]:Under the tide of big data, the Internet not only makes it convenient for people to obtain data, but also puts people in the dilemma of mass information choice. Although traditional search engines can filter out some irrelevant information, it is also time-consuming to select one by one from a large number of search results. Academia and industry agree that recommendation system is the most effective way to alleviate big data disaster. Music has become an important part of human life. With the development of Internet service, the consumption mode of music has changed greatly, and the user-centered technology has become the mainstream technology of music service. Social tags not only reflect the characteristics of resources, but also reflect the interests of users. At the same time, the extent to which users are interested in labels is constantly changing at different times. In this paper, a recommendation algorithm model based on label and time weighting is proposed. By using user behavior log, the user interest model under time migration is established, and the content of songs is identified by social tags, and then the recommendation list is generated according to the matching degree between user interest model and label characteristics of alternative songs. At the same time, the label can explain the reason why the project is recommended and improve the acceptance of the user. In order to solve the problem of data sparsity, we also strengthen the use of implicit data. In order to solve the cold start problem of new users, this paper proposes a solution for finding similar users based on user's basic information. Firstly, by using the similarity calculation rule derived from the material diffusion algorithm of the three-part graph, the users with similar behavior in the system are clustered, and then the basic information of the user is related to its category, and then the basic information of the new user can be based on the basic information of the new user. Find the similar user group, assign the initial label preference feature, and generate the personalized recommendation list for the new user. Finally, according to the requirements of the personalized music recommendation system, this paper designs the overall framework of the recommendation system, divides each functional sub-module, and designs and implements each module in detail. In the interactive module, the design of the music visualization interface can show the relationship between projects intuitively and improve the user experience.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2018
【分类号】：TP391.3
【参考文献】
相关期刊论文 前4条
1 杨卫芳;李学明;乔保学;;改进的热传导和物质扩散混合推荐算法[J];计算机工程;2017年03期
2 杨倩;潘兴德;;音乐推荐技术的现状与发展[J];电声技术;2012年06期
3 王国霞;刘贺平;;个性化推荐系统综述[J];计算机工程与应用;2012年07期
4 屈天喜;黄东军;童卡娜;;音乐可视化研究综述[J];计算机科学;2007年09期
，本文编号：2329780
本文链接：https://www.wllwen.com/kejilunwen/sousuoyinqinglunwen/2329780.html
