基于协同过滤算法的音乐推荐系统
发布时间：2018-09-15 06:29
【摘要】：伴随着近几年网络技术的不断普及发展，信息在互联网上的增长呈现爆炸式的速度，如此庞大的信息给用户带来了检索的困难，传统的搜索引擎技术难以满足用户的需求。在这种现有技术无法满足需要的背景下，产生了协同过滤系统，进而发展成为推荐领城的科研热点。 协同过滤算法主要分为基于内存的协同过滤算法和基于模型的协同过滤算法两大类。它是一种基于一系列兴趣相同的用户或项目进行的推荐，它参照相邻用户的偏好信息产生对目标用户的推荐清单。本文在比较了两者在数据集上的实验结果后选择采用基于项目的改正的余弦相似性算法作为我们音乐推荐系统的协同过滤算法。 该算法在一定程度上弥补了传统推荐方式的不足，同时本文结合协同过滤推荐算法带来的一定优势对现有的音乐推荐系统做了详细的分析，最后采用B/S软件结构实现了整个个性化音乐推荐系统的软件原型。 本文对音乐推荐原型系统进行了用户满意度的评测，并且对系统的可用性进行了测试。通过评估推荐结果我们发现该系统的整体性能较好，能够在一定程度上有效地推荐出用户可能感兴趣的结果，基本上达到了系统设计的预期目标。
[Abstract]:With the continuous development of network technology in recent years, the growth of information on the Internet presents an explosive speed, such a huge amount of information to users have brought difficulties in retrieval, the traditional search engine technology is difficult to meet the needs of users. Under the background that the existing technology can not meet the needs, collaborative filtering system has developed into a hot research spot in the recommended city. Collaborative filtering algorithms are divided into two categories: memory-based collaborative filtering and model-based collaborative filtering. It is a kind of recommendation based on a series of users or items of the same interest. It generates a list of recommendations to target users according to the preference information of neighboring users. In this paper, after comparing the experimental results of the two data sets, we choose the item-based correction cosine similarity algorithm as the collaborative filtering algorithm for our music recommendation system. To some extent, this algorithm makes up for the shortcomings of the traditional recommendation methods. At the same time, this paper makes a detailed analysis of the existing music recommendation system by combining the advantages of the collaborative filtering recommendation algorithm. Finally, the software prototype of the individual music recommendation system is realized by using the B / S software structure. In this paper, the user satisfaction of the music recommendation prototype system is evaluated, and the usability of the system is tested. Through the evaluation of the recommended results we find that the overall performance of the system is good and can effectively recommend the results that the users may be interested in to a certain extent and basically reach the expected goal of the system design.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3
本文编号：2244055
[Abstract]:With the continuous development of network technology in recent years, the growth of information on the Internet presents an explosive speed, such a huge amount of information to users have brought difficulties in retrieval, the traditional search engine technology is difficult to meet the needs of users. Under the background that the existing technology can not meet the needs, collaborative filtering system has developed into a hot research spot in the recommended city. Collaborative filtering algorithms are divided into two categories: memory-based collaborative filtering and model-based collaborative filtering. It is a kind of recommendation based on a series of users or items of the same interest. It generates a list of recommendations to target users according to the preference information of neighboring users. In this paper, after comparing the experimental results of the two data sets, we choose the item-based correction cosine similarity algorithm as the collaborative filtering algorithm for our music recommendation system. To some extent, this algorithm makes up for the shortcomings of the traditional recommendation methods. At the same time, this paper makes a detailed analysis of the existing music recommendation system by combining the advantages of the collaborative filtering recommendation algorithm. Finally, the software prototype of the individual music recommendation system is realized by using the B / S software structure. In this paper, the user satisfaction of the music recommendation prototype system is evaluated, and the usability of the system is tested. Through the evaluation of the recommended results we find that the overall performance of the system is good and can effectively recommend the results that the users may be interested in to a certain extent and basically reach the expected goal of the system design.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3
【参考文献】
相关期刊论文 前6条
1 赵亮,胡乃静,张守志;个性化推荐算法设计[J];计算机研究与发展;2002年08期
2 余力,刘鲁;电子商务个性化推荐研究[J];计算机集成制造系统;2004年10期
3 吴颜;沈洁;顾天竺;陈晓红;李慧;张舒;;协同过滤推荐系统中数据稀疏问题的解决[J];计算机应用研究;2007年06期
4 邓爱林,朱扬勇,施伯乐;基于项目评分预测的协同过滤推荐算法[J];软件学报;2003年09期
5 徐兰芳;胡怀飞;桑子夏;徐凤鸣;邹德清;;基于灰色系统理论的信誉报告机制[J];软件学报;2007年07期
6 马宏伟;张光卫;李鹏;;协同过滤推荐算法综述[J];小型微型计算机系统;2009年07期
相关博士学位论文 前2条
1 郭艳红;推荐系统的协同过滤算法与应用研究[D];大连理工大学;2008年
2 李涛;推荐系统中若干关键问题研究[D];南京航空航天大学;2009年
，本文编号：2244055
本文链接：https://www.wllwen.com/kejilunwen/sousuoyinqinglunwen/2244055.html
