The impact of diversity on clustering ensemble using Chi(2) criterion

Clustering ensemble is a technique for improving clustering results' robustness and accuracy. Basically, this technique generates base clusterings and then combines them into a consensus solution whose quality is determined by the diversity of the base clusterings and the consensus function's performance. In order to improve the quality of consensus solutions, it is necessary to generate base clusterings with regard to quality and diversity. Novel techniques were employed in this study to generate diverse base clusterings for both low-dimensional and high-dimensional datasets, as well as new criteria to compute the diversity of base clusterings with respect to quality. The impacts of different levels of diversity on consensus functions were studied. The proposed methods generated diverse base clusterings, according to the findings of the experiments.