基于竞争凝聚的个性化网页推荐

Personalized Web recommending based on competitive agglomeration

为了提高网站访问效率并得到有价值的个性化网页推荐,针对Web日志的新特性,提出了一种新的基于竞争凝聚的聚类算法.新算法对K-paths聚类算法进行了扩展和改进,按照路径的相似性进行聚类,采用竞争凝聚的思想,自动确定最佳的聚类数目.由于算法考虑了用户的访问兴趣,个性化网页推荐不打扰用户且不需要用户注册信息.利用关联规则得到个性化网页推荐集.用户推荐集和页面推荐集的结合大大提高了推荐效果,具有较好的扩展性.实验结果表明,与其他聚类方法相比该算法具有更高的推荐精度.

A new clustering algorithm based on competitive agglomeration was proposed to improve Web user＇s efficiency and get useful personalized Web recommending. As an improvement to the K-paths clustering algorithm, the new algorithm is a clustering algorithm according to path similarity, and can get best cluster numbers automatically by competitive agglomeration method. It does not disturb users and does not need any registration information because the algorithm takes into consideration the characteristics of user access sequence. The recommending system uses associate rules and integrates user clustering and page clustering to get recommending set of each user class. Experimental results showed that the correct rate of recommending was improved efficiently by using the new algorithm.