An effective modified possibilistic Fuzzy C-Means clustering algorithm for noisy data problems
Souad Azzouzi, Jaouad El-Mekkaoui, Amal Hjouji, Ahmed Elkhalfi · 2021
Clustering is a machine learning method that consists in grouping data points by similarity. Different Fuzzy C-Means clustering algorithms have been proposed, for example FCM, PCM, PFCM, most of those algorithms encounter several problems like choice of the adequate distance, efficiency against noise and outliers. In this study, we propose a new robust algorithm named Modified Possibilistic Fuzzy C-Means algorithm (MPFCM) based on possibilistic approach, which improves PFCM algorithm and overcomes those shortcomings. MPFCM is an extension of PFCM algorithm. Furthermore, the MPFCM allows to use more sophisticated norms for different complex problems. In another hand, MPFCM detects cluster centers more accurately in noisy data environment and with nonlinearly separable input data space. Experiments and results showed the effectiveness of our proposed algorithm MPFCM.