摘 要据统计,2015年间,数字音乐市场规模的增速大幅下降。同时其他数据表明,数字音乐消费仍有着巨大的潜在市场。由此可见通过智能推荐的方式来提升音乐产值是有必要的。目前的音乐智能推荐技术大致可以分为基于歌曲内容的算法、基于标签的算法以及基于流行度的算法。但这些算法或多或少存在着一些弊端。因此本文希望能通过关联分析解决数字音乐平台智能推荐歌曲的问题。
Apriori算法是挖掘布尔关联规则频繁项目集的算法中的一种。其通过不停地迭代来发现频繁项目集。算法在不能发现满足条件的频繁项目集时停止迭代。在实施过程中,Apriori算法需要根据实际情况被设定支持度。Apriori算法的优点在于:1、适用于稀疏的数据格式。2、通过频繁项目集的定义可以节省大量的运算时间。90490
本文运用Apriori算法在真实数据上完成关联规则的学习,并通过预留一部分测试数据的方式对得到的关联规则进行检验。检验结果证明依据本文算法所实现的推荐规则确有其可行性及实际应用价值。
Abstract According to statistics, the growth of digital music market rate dropped significantly in 2015。 While other data show that digital music consumption still has a huge potential market。 Therefore it is necessary to enhance the music output by means of intelligent recommendation。 The current music intelligent recommendation technology can be pided into Algorithm based on song content, tag-based algorithm, algorithm based on popularity。 However, these algorithms are more or less there are some drawbacks。 In this article, we wish to work out the problem of the digital music platform to recommend songs intelligently by affinity analysis。
The Apriori algorithm is one of the algorithms for mining Boolean association rules from frequent itemsets。 It finds frequent itemsets by constantly iterating。 The algorithm stops the iteration when the frequent item set satisfies the condition cannot be found。 In the implementation, Apriori algorithm needs to be set support according to the actual。 Apriori algorithm has the following advantages: 1。 Suitable for sparse data format。 2。 Through the definition of frequent itemsets can save a lot of computing time。
In this article, Apriori algorithm is used to complete the learning of association rules on real data, and the association rules are tested by reserving some of data。 The test results show that the recommended rules based on the algorithm are feasible and practical。
毕业论文关键词:歌曲推荐; 关联分析; 稀疏的矩阵格式; Apriori算法; Python
Keyword: Song Recommending; Affinity analysis; Sparse matrix format; The Apriori algorithm; Python
目 录
1。 绪论 4
1。1。 研究背景和意义 4
1。3。 研究内容和研究方法 5
2。 关联分析 5
2。1。 关联分析算法 5
2。2。 Apriori算法 5
2。3。 选择参数 6
3。 歌曲推荐问题 6
3。1。 获取数据集