A STEP TOWARDS THE MAJORITY-BASED CLUSTERING VALIDATION DECISION FUSION METHOD

A variety of clustering validation indices (CVIs) aimed at validating the results of clustering analysis and determining which clustering algorithm performs best. Different validation indices may be appropriate for different clustering algorithms or partition dissimilarity measures; however, the best suitable index to use in practice remains unknown. A single CVI is generally unable to handle the wide variability and scalability of the data and cope successfully with all the contexts. Therefore, one of the popular approaches is to use a combination of multiple CVIs and fuse their votes into the final decision. The aim of this work is to analyze the majority-based decision fusion method. Thus, the experimental work consisted of designing and implementing the NbClust majority-based decision fusion method and then evaluating the CVIs performance with different clustering algorithms and dissimilarity measures in order to discover the best validation configuration. Moreover, the author proposed to enhance the standard majority-based decision fusion method with straightforward rules for the maximum efficiency of the validation procedure. The result showed that the designed enhanced method with an invasive validation configuration could cope with almost all data sets (99%) with different experimental factors (density, dimensionality, number of clusters, etc.).

Clustering assessment in weighted networks

We provide a systematic approach to validate the results of clustering methods on weighted networks, in particular for the cases where the existence of a community structure is unknown. Our validation of clustering comprises a set of criteria for assessing their significance and stability. To test for cluster significance, we introduce a set of community scoring functions adapted to weighted networks, and systematically compare their values to those of a suitable null model. For this we propose a switching model to produce randomized graphs with weighted edges while maintaining the degree distribution constant. To test for cluster stability, we introduce a non parametric bootstrap method combined with similarity metrics derived from information theory and combinatorics. In order to assess the effectiveness of our clustering quality evaluation methods, we test them on synthetically generated weighted networks with a ground truth community structure of varying strength based on the stochastic block model construction. When applying the proposed methods to these synthetic ground truth networks’ clusters, as well as to other weighted networks with known community structure, these correctly identify the best performing algorithms, which suggests their adequacy for cases where the clustering structure is not known. We test our clustering validation methods on a varied collection of well known clustering algorithms applied to the synthetically generated networks and to several real world weighted networks. All our clustering validation methods are implemented in R, and will be released in the upcoming package clustAnalytics.