Correlated-clustering Frame A Holistic Method of Deep Web Sc

Matching Based on Data MiningFu Yuchen, Liu Quan, Xu Yunlong, Zhang Chao, Zhou Wenyun, Cui ZhimingSchool of Computer Science and TechnologySoochow UniversitySuzhou, Jiangsu Province, China 215006yuchenfu@AbstractA large number of Deep Web data sources are onlyaccessible through their query interfaces. For anydomain of interest, there may be many such sourceswith varied query capabilities and content coverage.To obtain mass valuable information in deep Web, weneed to integrate large heterogeneous information.Schema matching is a critical problem in theintegration process. This paper propose a new holisticschema matching method based on data mining,named as Correlated-clustering, which minespositively correlated attributes to form potentialattribute groups, and finds synonym attributes byclustering. we design experiments to implementmentioned algorithms and technology. Experimentalresults testify that our solution achieves accurately andeffectively.1. IntroductionIn Deep Web, online databases provide the dynamicquery-based data access method through the queryinterface, each query interface accepts queries throughtheir query schema (for example, the author, title,theme ...). Query schema of the Deep Web usuallygenerates complex matches between the propertygroups. Compared with 1:1 match, complex match isbuild a association between a property group with a setof M attributes and another property group with a setof N attributes, i.e. m:n match. For example, in thebook domain, {author}={first name, last name}; in the