Towards for Using Spectral Clustering in Graph Mining

This paper presents an approach of community detection from data modeled by graphs, using the Spectral Clustering (SC) algorithms, and based on a matrix representation of the graphs. We will focus on the use of Laplacian matrices afterwards. The spectral analysis of those matrices can give us interesting details about the processed graph. The input of the process is a set of data and the output will be a set of communities or clusters that regroup the input data, by starting with the graphical modeling of the data and going through the matrix representation of the similarity graph, then the spectral analysis of the Laplacian matrices, the process will finish with the results interpretation.