Comparative Study of Clustering Algorithms using OverallSimSUX Similarity Function for XML Documents

A huge amount of information is represented in XML format. Several tools have been developed to store, and query XML data. It becomes inevitable to develop high performance techniques for efficiently analysing extremely large collections of XML data. One of the methods that many researchers have focused on is clustering, which groups similar XML data, according to their content and structures. In previous work, there has been proposed the similarity function OverallSimSUX, that facilitates to capture the degree of similitude among the documents with a novel methodology for clustering XML documents using both structural and content features. Although this methodology shows good performance, endorsed by experiments with several corpus and statistical tests, on having had impliedly only one clustering algorithm, K-Star, we do not know the effect that it would suffer if we replaced this algorithm by other with dissimilar characteristics. Therefore to endorse completely the methodology, in this work we make a comparative study of the effects of applying the methodology for the OverallSimSUX similarity function calculation, using clustering algorithms of different classifications . Based on our analysis, we arrived to two important results: (1) The Fuzzy-SKWIC clustering algorithm works best both with methodology and without methodology, although there are not present significant differences respect to the K-Star clustering algorithm; (2) For each analysed algorithm when using the methodology, we obtain better results than when it is not taken into account.