Clustering with Minimum Spanning Tree using TOPSIS with Multi-Criteria Information
José G. M. Esgario, Renato Antonio Krohling · 2018
Clustering is the process of grouping similar objects into the same partition and keeping dissimilar objects on different partitions. Clustering algorithms based on Minimum Spanning Tree (MST) have been successfully applied in the separation of non-convex clusters, although the use of individual objective functions limits the algorithms to a reduced set of clustering problems, presenting difficulties in cases of unbalanced, noisy, overlapping datasets, etc. In order to make clustering process more robust, this paper proposes an algorithm based on Minimum Spanning Tree that combines different objective functions using TOPSIS. The algorithm performance was evaluated on real and synthetic datasets. Experimental results indicate that the combination of objective functions improves clustering results compared to individual functions.