A Self Adaptive FCM Cluster Forests Based Feature Selection

Ensemble clustering refers to combine many clustering methods to produce better results. In this context, we propose a new clustering ensemble method inspired from cluster forests (CF) based Self-Adaptive Fuzzy C-Means (SAFCM) method. Firstly, unsupervised feature selection methodology based on the building of best variables on simulated datasets. Next, we ameliorate the CF algorithm with the integration of SAFCM to find also the best number of K groups. Finally, the modified version normalized cuts spectral clustering (Ncut) is applied to general grouping. The proposed algorithm was tested on datasets from UCI Machine Learning Repository. The experimental results indicate that our proposed method outperforms both different clustering algorithms in terms of clustering quality.

I. Introduction

Clustering algorithms is a fundamental approach to analyze correctly information from the data analytics large social. It aims to separate the data points into various clusters according to a certain similarity measure [1]. Clustering algorithms are categorized as: crisp and fuzzy ones. In crisp methods, the membership degree of each data element is zeros or one. In fuzzy methods, data elements are assigned to many cluster with a membership degree between zeros and one.