基于DB-CF算法的音乐平台个性化推荐研究

Research on Personalized Recommendation of Music Platform Based on DB-CF Algorithm

娱乐方式日益丰富,产生巨量数据,利用这些数据通过推荐系统可以让用户获得更好的体验,为此提出了DB-CF(DBSCAN-Collaborative Filtering)算法。首先,使用DBSCAN聚类算法对音乐平台的线下用户进行聚类;然后,通过协同过滤算法计算对象用户与各聚类中心的相似度,再通过对比相似度度量矩阵,遍历离对象用户最近的邻居,通过邻居作出评分预测。实验表明,采用DB-CF算法比传统算法准确率提高8%左右,可以产生更准确的推荐结果,为用户带来更好的体验。

With the increasing enrichment of entertainment methods and the influx of huge amounts of data,people’s lives are more convenient through the effective use of data.In terms of music platforms,excellent recommendation systems are used to provide better experience to platform users.In order to obtain more accurate recommendation results than the traditional recommendation techniques in a large number of tracks,a DBSCAN-collaborative filtering(DB-CF)algorithm is proposed.Firstly,when processing the offline data,we use the DBSCAN clustering algorithm to cluster the users of the music platform.Secondly,when processing the online data,we calculate the similarity between the user and each cluster center by a collaborative filtering algorithm.And then through the comparison of the similarity measurement matrix,we traverse the nearest neighbor of the object user,and make prediction of the object user’s score.Experiments show that under different recommendation algorithms,the DB-CF algorithm improves the accuracy by about 8%compared with the traditional algorithm,which proves the algorithm can produce more accurate recommendation results and bring better experience to users.