Kuwil method for spectral clustering algorithm

Abstract

The open issues and challenges that exist while using the spectral clustering algorithm (SCA) have led to its limited spread in practical life. This paper proposes to find an easier, faster and more accurate method to implement SCA that will lead to its wide use by statisticians, researchers, institutions and others. I suggest a new method called ‘Kuwil method’ for SCA on any dataset points without needing estimation or evaluation of any parameters or the use of linear algebra, not even the k_mean algorithm. The main aim is to apply an algorithm that relies on distance laws among points only. The algorithm by the Kuwil method has been applied a number of times on real data from the warehouse European Economic Association (http://ec.europa.eu/eurostat/data/database) and on unreal data. The results were highly efficient in terms of time, effort and simplification. It eliminates the problem of parameters and increases the effectiveness to give static results obtained from the first execution. No errors were seen from functions in the MATLAB language such as eigenvalues, eigenvector and k_mean.

Keywords: Spectral clustering, Kuwil method.