Towards a scalable approach for mining frequent patterns from the linked open data cloud

Abstract

In recent years, the linked data principles have become one of the prominent ways to interlink and publish datasets on the web creating the web space a big data store. With the data published in RDF form and available as open data on the web opens up a new dimension to discover knowledge from the heterogeneous sources. The major problem with the linked open data is the heterogeneity and the massive volume along with the preprocessing requirements for its consumption. The massive volume also constraint the high memory dependencies of the data structures required for methods in the mining process in addition to the mining process overheads. This paper proposes to extract and store the RDF dumps available for the source data from the linked open data cloud which can be further retrieved and put in a format for mining and then suggests the applicability of an efficient method to generate frequent patterns from these huge volumes of data without any constraint of the memory requirement. © Springer International Publishing Switzerland 2014.