Özet:

Clustering is an important method for customer segmentation, and there are lots of clustering methods in literature and most of them are similarity-based. However, there are also some remarkable dissimilarity-based approaches that should be compared with others to analyze their practical benefits. In this study, we apply two different clustering techniques to figure out what differences can be observed between similarity and dissimilarity-based clustering techniques. To provide this, we have conducted a study on customer database with 8000 customers from automotive industry. We have chosen self-organizing maps, the most popular and acceptable similarity-based method, and dissimilarity-based heuristic method that is using game theory and evolutionary algorithm. Experimental results show that dissimilarity-based heuristic method has achieved better performance in determining more specified clusters and these two methods should be used for different goals because of the final clusters specifications. At the same time, a successful example of new emerging method using game theory in clustering has been completed.

Anahtar Kelimeler: