TY - JOUR

T1 - Evaluation and selection of clustering methods using a hybrid group MCDM

AU - Barak, S.

AU - Mokfi, T.

N1 - This is the author’s version of a work that was accepted for publication in Expert Systems with Applications. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Expert Systems with Applications, 138, 2019 DOI: 10.1016/j.eswa.2019.07.034

PY - 2019/12/30

Y1 - 2019/12/30

N2 - Due to the lack of objective measures, the evaluation and prioritization of clustering methods is inherently challenging. Since their evaluation generally involves numerous criteria, it can be designed as a multiple criteria decision making (MCDM) problem and using multiple data sets, the problem can be formulated as a group MCDM modeling.In this paper, a MCDM-based framework is proposed to evaluate and rank a number of clustering methods. The proposed approach employs three group MCDM algorithms and a Borda count method which leads to a comprehensive, robust framework capable of evaluating and ranking multiple clustering models on manifold data sets (cases). Moreover, we introduce a hybrid data clustering algorithm which combines a particle swarm optimization (PSO) algorithm with a K-means clustering algorithm. Finally, a clustering comparison with regard to both external and internal evaluation indicators is another contribution of this paper.Six clustering methods are compared based on five evaluation measures. The results of comparative experiments on ten data sets indicate the effectiveness of the proposed hybrid clustering method. More importantly, the experimental results vividly demonstrate the effectiveness of the group MCDM-based evaluation on clustering model selection.

AB - Due to the lack of objective measures, the evaluation and prioritization of clustering methods is inherently challenging. Since their evaluation generally involves numerous criteria, it can be designed as a multiple criteria decision making (MCDM) problem and using multiple data sets, the problem can be formulated as a group MCDM modeling.In this paper, a MCDM-based framework is proposed to evaluate and rank a number of clustering methods. The proposed approach employs three group MCDM algorithms and a Borda count method which leads to a comprehensive, robust framework capable of evaluating and ranking multiple clustering models on manifold data sets (cases). Moreover, we introduce a hybrid data clustering algorithm which combines a particle swarm optimization (PSO) algorithm with a K-means clustering algorithm. Finally, a clustering comparison with regard to both external and internal evaluation indicators is another contribution of this paper.Six clustering methods are compared based on five evaluation measures. The results of comparative experiments on ten data sets indicate the effectiveness of the proposed hybrid clustering method. More importantly, the experimental results vividly demonstrate the effectiveness of the group MCDM-based evaluation on clustering model selection.

KW - Clustering

KW - MCDM

KW - Group TOPSIS

KW - Group COPRAS

KW - Particle Swarm Optimization

U2 - 10.1016/j.eswa.2019.07.034

DO - 10.1016/j.eswa.2019.07.034

M3 - Journal article

VL - 138

JO - Expert Systems with Applications

JF - Expert Systems with Applications

SN - 0957-4174

M1 - 112817

ER -