Mining Frequently Changing Substructures from Historical Unordered XML Documents

Recently, there is an increasing research efforts in XML data mining. These efforts largely assumed that XML documents are static. However, in many real applications, XML data are evolutionary in nature. In this paper, we focus on mining evolution patterns from historical XML documents. Specifically, we propose a novel approach to discover frequently changing structures (FCS) from a sequence of historical versions of unordered XML documents. The objective is to extract substructures that change frequently and significantly by analyzing structural evolution patterns of XML documents. We propose two algorithms based on a set of evolution metrics to extract FCS from the historical XML data. We also present a battery of optimization techniques to improve the space efficiency of our algorithms. Note that such structures cannot be extracted accurately and efficiently by repeatedly applying existing frequent substructure mining techniques on a sequence of snapshot data. FCS can be useful in several applications such as monitoring interesting structures in a specific domain, FCS-based classifier, indexing XML documents, and evolution-conscious XML query caching. Extensive experiments with both synthetic and real data show that the proposed algorithms are efficient and scalable and can discover FCS accurately.