A Shortest Path Similarity Matrix based Spectral Clustering

This paper proposed a new spectral graph clustering model by casting the non-categorical spatial data sets into an undirected graph. Decomposition of the graph to Delaunay graph has been done for computational efficiency. All pair shortest path based model has been adapted for the creation of the underlying Laplacian matrix of the graph. The similarity among the nodes of the graph is measured by a random selection based correlation coefficients. The effectiveness as well as the efficiency of the proposed model has beentasted and measuredwith standard data and the performances are compared with that of existing standard models. Keywords—Graph Clustering;Delaunay Triangulation;All-pair Shortest Path Distance; Similarity Matrix; Spectral Clustering.