Evolutionary, Neural and Statistical Approaches to interval Clustering for Web Mining

 

Pawan Lingras, Rui Yan & Mofreh Hogo

 

Department of Math and Computer Science, Saint Mary’s University

Halifax, Nova Scotia, Canada, B3H 3C3

Department of Computer Science and Engineering, Faculty of Electrical Engineering, Czech Technical University, Karlovo Nam. 13, 121 35 Prague 2, Czech Prepublic

 

Abstract

 

Rough set theory has been used extensively for supervised learning. Clustering within the context of rough set theory is attracting increasing interest among researchers. Typical clustering operations in data mining involve finding natural groupings of resources or users. Conventional clusters have crisp boundaries, i.e. each object only belongs to only one cluster. The clusters and associations in data mining do not necessarily have crisp boundaries. An object may belong to more than one cluster. This paper describes three different methodologies based on properties of rough sets for developing interval representations of clusters. The first approach is based on Genetic Algorithms, the second approach is an adaptation of the K-means algorithm, and the last is based on Kohonen self-organizing maps. The paper also provides an experiment to illustrate the rough set based clustering of web users.

 

Keywords

 

Clustering, rough sets, Kohonen self-organizing maps, K-means, evolutionary computing