基于模糊聚类的可扩展的协同过滤方法  

Collaborative filtering method with enhancing scalability using fuzzy clustering

作　　者：王晓军[1] 付超[2] 

机构地区：[1]南京邮电大学物联网学院,江苏南京210003 [2]南京邮电大学计算机学院,江苏南京210023

出　　处：《南京邮电大学学报（自然科学版）》2017年第1期14-18,共5页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61003237)资助项目

摘　　要：作为电子商务系统领域里一种重要的推荐算法,协同过滤在过去十年取得了广泛的传播和使用。但随着数据库中用户和商品数量的增长,为用户做出推荐所需的计算成本逐渐加大,可扩展性成为协同过滤面临的重要挑战。为了改善协同过滤的可扩展性,提出一种基于模糊聚类的可扩展的协同过滤方法。该方法首先根据项目特征利用模糊聚类算法对项目进行聚类,并在簇内产生潜在相似关系集合,然后将潜在相似关系集合进行分区并在各个分区内并行计算项目的相似度,最后搜索邻居并做出推荐。实验结果表明,此方法在提高协同过滤推荐系统的可扩展性上取得了一些好的结果,同时保持了较好的精确度。As an important recommendation algorithm in the field of e-commerce system, collaborative fil- tering has been widely used in the past ten years. However, with the increase of the number of users and goods, the computational cost for making recommendations is gradually increasing, and the scalability be- comes an important challenge for the collaborative filtering. To improve the scalability of collaborative fil- tering, a scalable collaborative filtering method based on fuzzy clustering is proposed. Firstly, clustering items are made by fuzzy K-means algorithm according to the item characteristics, and a potential similar relationship is produced within the clusters. Then, the potential similar relationship set is partitioned and the item similarity in the corresponding partitions is parallel computed. Finally, the neighbors are searched to make recommendations . Experimental results show that the proposed method achieves good results in improving the scalability of collaborative filtering recommender systems, while maintaining good accuracy.

关 键 词：推荐系统 协同过滤 扩展性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]