K-Means Clustering of Ontologies Based on Graph Metrics

The growth of the semantic web in recent years have led to an increase in the development of ontologies. The increase in the number of ontologies requires users or developers to understand their design for the purposes of selection and reuse. On the other hand, the availability of various ontology metrics allows for different methods of analyzing and evaluating the ontologies. In this paper, the K-Means clustering algorithm have been implemented on a set of graph metrics for analyzing ontologies. The experiments were carried out with k values from 2 to 7, resulting to 2, 3, 4, 5, 6 and 7 clusters, respectively. The clustering results were further visualized with scatterplots. The scatterplots enabled to visualize the effect of various k values on the centroids of the clusters and to conclude that the higher the k is, the closer the centroids of the clusters are to the data points. The experiments showed that the K-Means algorithm is efficient in clustering ontologies based on their graph metrics.