基于MDS变换聚类算法的数字电视用户特征提取

The Extraction of Digital TV User Feature Based on MDS Clustering

提出利用MDS(Multidimensional Scaling)变换聚类算法提取数字电视用户的收视特征,解决了传统聚类算法因中间聚类中心无距离度量而无法应用的问题.基于实际运行的有线数字电视系统,建立了由时间、频道、节目主类别、节目子类别表述的节目特征模型及节目差异模型;提出了基于MDS聚类算法提取用户收视特征的具体步骤;基于实际用户收视记录的计算结果具有特征一致性,以提取特征为基准的节目推荐结果与用户实际的收视记录比对,具有70%准确性.

A MDS clustering algorithm using for extraction of digital TV user feature is proposed, and solved the problem that the classic clustering algorithm can not be used because of the TV program sample having no distance measurement. The program feature model and the program difference model expressed by time,channel,main category and subcategory are founded. The detail process of MDS clustering algorithm using to pick up TV user feature is presented. The calculated feature based on real user watching record conforms real user action well with 70% orecision.