TR 00-034 A Comparison of Document Clustering Techniques

This paper presents the results of an experimental study of some common document clustering techniques. In particular, we compare the two main approaches to document clustering, agglomerative hierarchical clustering and K-means. (For K-means we used a “standard” K-means algorithm and a variant of K-means, “bisecting” K-means.) Hierarchical clustering is often portrayed as the better quality clustering approach, but is limited because of its quadratic time complexity. In contrast, K-means and its variants have a time complexity which is linear in the number of documents, but are thought to produce inferior clusters. Sometimes K-means and agglomerative hierarchical approaches are combined so as to “get the best of both worlds.” However, our results indicate that the bisecting K-means technique is better than the standard K-means approach and as good or better than the hierarchical approaches that we tested for a variety of cluster evaluation metrics. We propose an explanation for these results that is based on an analysis of the specifics of the clustering algorithms and the nature of document data. 1 Background and Motivation Document clustering has been investigated for use in a number of different areas of text mining and information retrieval. Initially, document clustering was investigated for improving the precision or recall in information retrieval systems [Rij79, Kow97] and as an efficient way of finding the nearest neighbors of a document [BL85]. More recently, clustering has been proposed for use in browsing a collection of documents [CKPT92] or in organizing the results returned by a search engine in response to a user’s query [ZEMK97]. Document clustering has also been used to automatically generate hierarchical clusters of documents [KS97]. (The automatic generation of a taxonomy of Web documents like that provided by Yahoo! (www.yahoo.com) is often cited as a goal.) A somewhat different approach [AGY99] finds the