Cluster validity measurement for arbitrary shaped clusters
Ferenc Kovács, Renáta Iváncsy · International Conference on Artificial Intelligence · 2006
Clustering is an unsupervised process in data mining and pattern recognition and most of the cluster- ing algorithms are very sensitive to their input parameters. Therefore it is very important to evaluate the result of the clustering algorithms. In this pater a novel validity measurement index is introduced which can be used for evaluating arbitrary shaped clusters. The main advantage of this validity index are the following: it can compare and measure not only elliptical clusters but arbitrary shaped clusters as well. This validity index can evaluate the result of any clustering algorithms but the calculation of this index is very simple in case of density based algorithms as it can be calculated during the clustering process.