A multiple kernels interval type-2 possibilistic C-means

In this paper, we propose multiple kernels-based interval type-2 possibilistic c-Means (MKIT2PCM) by using the kernel approach to possibilistic clustering. Kernel-based fuzzy clustering has exhibited quality of clustering results in comparison with “routine” fuzzy clustering algorithms like fuzzy c-Means (FCM) or possibilistic c-Means (PCM) not only noisy data sets but also overlapping between prototypes. Gaussian kernels are suitable for these cases. Interval type-2 fuzzy sets have shown the advantages in handling uncertainty. In this study, multiple kernel method are combined into interval type-2 possibilistic c-Means (IT2PCM) to produce a variant of IT2PCM, called multiple kernels interval type-2 possibilistic c-Means (MKIT2PCM). Experiments on various data-sets with validity indexes show the performance of the proposed algorithms. © Springer International Publishing Switzerland 2016.