Anti-noise possibilistic clustering based on maximum entropy
Xingguang Pan, Xiongtao Zhang, Zhibin Jiang, Shitong Wang · 2017
Maximum Entropy Clustering (MEC) is an algorithm based on fuzzy c means by embedding an entropy generalization term in it. However, MEC is not robust to both noise and outliers, which leads to poor accuracy in clustering processes. In this paper, a novel clustering algorithm based on Shannon entropy is proposed, the new algorithm named Anti-noise Possibilistic Maximum Entropy Clustering (A-PMEC) is verified much more robustness in noisy dataset. We introduce the detailed formulation of A-PMEC and as well as experimental study to demonstrate the merits of the proposed method.