一个基于Web访问路径聚类的智能推荐系统

提出了一个基于Web用户访问路径聚类的智能推荐系统．系统使用基于代理技术的结构，由离线的数据预处理和基于用户访问路径的URL聚类以及在线推荐引擎两部分组成．提出了一个基于用户浏览兴趣的推荐规则集生成算法，在度量用户浏览兴趣时综合考虑了用户浏览时间和对该页面的访问次数．提出了一个基于推荐规则集和站点URL路径长度的URL推荐算法．实验表明，该算法比使用基于关联规则和基于用户事务的推荐算法的精确性有较大幅度的提高．

An Intelligent Recommendation System Based on Web Navigation Path Clustering

An intelligent recommendation system is proposed based on clustering of Web user's navigation path.The system uses an architecture based on proxy techniques,and consists of two subsystems,i.e.,the offline subsystem,including data preparation and URL clustering based on user's browsing paths,and the online subsystem,including a recommendation engine and a Web HTTP server.A algorithm for generating recommendation rule set is proposed based on the user's browsing interest,which is measured by considering synthetically both the user's browsing time and the number of hits on the Web page.A recommendation algorithm is presented based on recommendation rule set and the length of Web site URLs.The experiments show that,comparing with the recommendation algorithms based on association rule or on user transaction,the algorithm precision is improved greatly.