Fast and effective clustering of XML data using structural information

This paper presents the incremental clustering algorithm, XML documents Clustering with Level Similarity (XCLS), that groups the XML documents according to structural similarity. A level structure format is introduced to represent the structure of XML documents for efficient processing. A global criterion function that measures the similarity between the new document and existing clusters is developed. It avoids the need to compute the pair-wise similarity between two individual documents and hence saves a huge amount of computing effort. XCLS is further modified to incorporate the semantic meanings of XML tags for investigating the trade-offs between accuracy and efficiency. The empirical analysis shows that the structural similarity overplays the semantic similarity in the clustering process of the structured data such as XML. The experimental analysis shows that the XCLS method is fast and accurate in clustering the heterogeneous documents by structures.