Automatic concept clustering for ontological structure through data mining techniques

Ontology is a technique for expressing formal specification. It is a conceptualization of domain and its terms and relationships. The technique of ontology finds its application in almost every area, some of which includes medicine, e-commerce, chemistry, education etc. Concept clustering is the foremost step in construction of ontology. Concept clustering is usually a manual process involves labor and time intensive task. Hence there is a need for automatic grouping of concepts for ontology construction. In this paper, automatic concept clustering is attempted through data mining clustering techniques. The training set for the concepts formation of ontology structure is obtained from zoo dataset in UCI Machine Learning Repository. The clustering techniques are implemented through Weka 3.7.6, an open source data mining tool. Performance of clustering techniques viz., EM, Farthest First and K-Means are analyzed. It is found that Farthest-First clustering technique yielded the best performance with an accuracy of 93.0693%. The methodology proposed in this paper can be adopted for any other case study also.