Determining the Number of Clusters with the TOPSIS Method in Clustering Based on the Multivariate Mixtures of Normal Distributions

44933

Multiple criteria decision making methods provide the selection, ordering and classification of the alternative among the possible solution sets with the optimization of multiple criteria. The aim of this study is to determine the number of clusters in model-based cluster analysis with the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution), which is one of the multiple criteria decision making methods. In the study, the data sets were modeled with model-based clustering according to the number of candidate clusters, and for each cluster obtained, the Akaike information criterion, the approximate weight criterion of evidence, Bayesian information criterion, classification likelihood criterion and Kullback information criterion is calculated as a decision criterion. Simulation results were used in weighting the criteria and the most suitable number of clusters was determined with the TOPSIS. The success of the proposed approach was tested on the real and the synthetic datasets. As a result of the application, the proposed approach in determining the appropriate cluster number performed better than the relevant information criteria.

472

Information Criteria , Multi Criteria Decision Making , Model Based Clustering , TOPSIS

Erciyes University Journal Of The Institute Of Science an‎d Technology

480

Erciyes University Journal Of The Institute Of Science an‎d Technology