中文摘要：

海量、异构的Web日志同时蕴含有巨大潜在价值的信息,为有效发现这些资源,并用于为用户提供更高效的服务,建立了智能电子商务模型,指出通过Web使用挖掘,发掘规律、模式和知识来支撑电子商务的智能。提出了细化最大向前引用模型算法,用于处理Web日志,获取用户事务序列,并与最大向前引用模型算法进行了比较,说明细化后的算法更能反映用户浏览习惯。将这些用户事务序列转换为二进制向量,并结合改进蚁群聚类算法,对它们进行了聚类操作,实现了用户簇聚。最后,建立了在线自动聚类的智能电子商务系统原型,并应用到了实际运营的Web系统中,验证了原型的合理性。

英文摘要：

Web logs are massive,heterogeneous and contain huge potential value. To effectively mine these resources so as to provide more efficient services for users,intelligent e-commerce model was set up,and laws,patterns and knowledge were acquired by Web usage mining to support e-commerce intelligence. Refining maximal forward reference model algorithm was presented to deal with Web log. User access transaction sequences were obtained by the model. Compared to maximal forward reference model,refining model could reflect users＇ Web-browse habits easily. These sequences were converted to binary vectors,and these vectors were clustered by extended ant colony clustering algorithm. Finally,online automatic cluster intelligent e-commerce prototype system was established. It was applied to commercial website,and the application result demonstrated the rationality of this prototype.