Generalized intuitionistic fuzzy c-means clustering algorithm using an adaptive intuitionistic fuzzification technique

Abstract

In real-world scenario, mostly, the datasets are either imprecise or uncertain in their original form. Due to this reason, the clustering of such datasets is unsatisfactory and we often get compromised results. The information present in the dataset can be made precise and useful with the popular generalization of fuzzy sets known as Atanassov intuitionistic fuzzy sets (AIFSs). Therefore, the AIFS-based c-means clustering algorithm becomes a convenient way for clustering uncertain and vague datasets. In the paper, we propose an intuitionistic fuzzy-based algorithm, namely Generalized Intuitionistic Fuzzy c-Means (G-IFCM) clustering algorithm which uses an adaptive AIFS Euclidean distance measure in its criterion function to cluster the dataset under intuitionistic fuzzy environment. The proposed intuitionistic fuzzification method incorporates a technique to transform the dataset into AIFS and maintains its original structure that tends to change during any fuzzification process. Further, simulation experiments are also conducted on few UCI machine learning repository datasets using G-IFCM and its performance is compared with some known clustering algorithms. The efficacy of the algorithm is tested with some popular benchmark indexes that check the cluster validity and clustering performance of G-IFCM. Finally, the computation suggests efficient performance of G-IFCM where the best choice of fuzzy factor, m probably lies in the interval [1, 2] that indicates reduction in the expense of computation.