Service-oriented architectures are widely adopted to implement distributed data analysis applications that exploit decentralized data mining services. Standard Web service (WS) technologies are often used to provide access to such services, in order to facilitate their composition and to effectively handle heterogeneity of data mining algorithms and implementations. In this paper we propose a framework for efficiently discovering data mining services in large-scale networks, taking into account the specific features of the algorithms and implementations of such services. The proposed framework, named WS-Chord, includes a taxonomy of data mining services used to annotate the WSs, and an extension of the Chord DHT protocol to efficiently index and discover annotated WSs. A locality preserving hashing (LPH) function, which takes into account the taxonomy hierarchy, has been defined to support efficient service discovery over WS-Chord when queries are formulated against the proposed taxonomy. Experimental results have demonstrated the effectiveness of the framework in terms of relevant data mining services discovered. © 2014 IEEE.
Add the full text or supplementary notes for the publication here using Markdown formatting.