Applied Analysis and Optimization, cilt.4, ss.191-206, 2020 (Diğer Kurumların Hakemli Dergileri)
Clustering is one of the main areas in data mining. The purpose of
this problem is to find clusters on unlabeled data such that within-cluster similarities are minimum. Although clustering problem is defined on unlabeled data,
we have the knowledge that some data should be in the same group in real life.
This information can be used for more useful solution of the clustering problem.
Such a problem is called semi-supervised clustering problem. In this study, two
incremental algorithms are proposed for effective solution of clustering problem.
We design the algorithms that based an incremental approach to generate good
starting points for obtaining cluster centers. Computational experiments have
been made on real datasets. The results of the proposed algorithms are competitive with the other methods in the literature and often suggest much better
objective function values.