基于蚁群算法与K-means算法相结合的Web用户聚类

An Integration of Ant Colony Optimization and K-means for Web Users Clustering

Web用户聚类是指用聚类算法产生用户会话的聚类，是电子商务中的一个重要问题。该问题的难度在于有成千上万的会话需要聚类，而且每个会话都可描述为一个高维向量。此外，该问题就聚类的数目而言具有指数的复杂性，是一个NP-难的问题。本文提出一种新的聚类方法，该方法将蚁群算法与K-means算法相结合对用户会话进行优化聚类。实验结果表明，与K—means算法相比，该方法在Web导航推荐的应用中具有更好的性能。

Web user clustering is an important issue in E-commerce. The clustering algorithm should create clusters of user sessions. This problem is difficult because thousands of sessions may have to be clustered, and also because each session is described for instance as a vector with high dimensionality. Moreover, this problem possesses exponential complexity in terms of number of clusters and become an NP-hard problem. This paper presents a novel methodology combining ant colony optimization with K-means for optimally clustering user sessions. Experimental results show that our approach is better than using only K-means in the application of Web navigation recommendation.