基于多视角二分k-means的高校图书馆用户画像研究  

Research on user portraits of university libraries based on multi-view binary k-means

作　　者：李伟[1] 胡云飞 李澎林[1] LI Wei;HU Yunfei;LI Penglin(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学计算机科学与技术学院,浙江杭州310023

出　　处：《浙江工业大学学报》2020年第2期141-147,共7页Journal of Zhejiang University of Technology

基　　金：浙江省教育厅基金资助项目(Y201840640)。

摘　　要：针对高校图书馆无法实现精准读者推荐和服务的问题,在充分分析读者在图书馆的行为数据基础上,设计了一种基于多视角聚类的高校图书馆用户画像框架。考虑到经典k-means算法在多视角聚类中存在容易陷入局部最优的缺陷,提出了一种基于马氏距离的多视角二分k-means算法,该算法引入马氏距离有效地解决了欧式距离在多视角聚类中受属性量纲的影响。实验证明:相比经典k-means算法和二分k-means算法,改进后的算法在用户画像过程中全局最优、鲁棒性好、效率高;利用该框架得到的用户画像能够帮助高校图书馆挖掘读者需求、提高服务水平。Aiming at the problem that college library can not achieve accurate recommendation and service for readers,a user portrait framework of university libraries based on multi-view clustering is designed on the basis of fully analyzing the behavior data of readers in libraries.Considering that the classical k-means algorithm is easy to fall into the local optimality in multi-view clustering,a multi-view binary k-means algorithm based on Mahalanobis distance is proposed.This algorithm introduces Mahalanobis distance to effectively solve the problem that Euclidean distance is affected by attribute dimension in multi-view clustering.Experiments show that compared with the classical k-means algorithm and the binary k-means algorithm,the improved algorithm is globally optimal,robust and efficient in the process of user portrait.User portraits obtained by using this framework can help university libraries to mine readers’needs and improve service levels.

关 键 词：数据挖掘 用户画像 行为分析 二分k-means算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]