Existing works on XML data mining deal with snapshot XML data only, while XML data is dynamic in real applications. In this paper, we discover knowledge from XML data by taking account its dynamic nature. We present a novel approach to extract pattern-based dynamic structures from versions of unordered XML documents. With the proposed dynamic metrics, the pattern-based dynamic structures are expected to summarize and predict interesting change trends of certain structures based on their past behaviors. Two types of pattern-based dynamic structures, increasing dynamic structure and decreasing dynamic structure are considered. With our proposed data model, SMH-Tree, an algorithm for mining such pattern-based dynamic structures with only two scans of the XML sequence is presented. Experimental results show that the proposed algorithm can extract the pattern-based dynamic structures efficiently with good scalability.
展开▼