Clustering research using dynamic modeling based on granular computing

摘要

Clustering techniques is a discovery process in data mining, especially used in characterizing customer groups based on purchasing patterns, categorizing Web documents, and so on. Many of the traditional clustering algorithms falter when the dimensionality of the feature space becomes high, relativing to the size of the document space, So it is important to precondition for modeling. Secondly, we are usually disappointed to their clustering speed. when having very large complex data sets, and another defect is that they always fit some static model, so if the user doesn't select appropriate static-model parameters, these algorithms can break down. In this paper, we introduce a new clustering algorithm to improve the speed of clustering, this clustering technique, which is based on granular computing and hypergraph partition, and it is capable of automatically discovering document similarities or associations, and this approach considers the internal characteristics of the clusters themselves, thus it doesn't depend on a static model. Finally, we conduct several experiments on real Web data searched by ordinary search engine and received the satisfied results.