An Insensitivity Fuzzy C-means Clustering Algorithm Based on Penalty Factor

This paper analyzes sensitivity of Fuzzy C-means to noisy which generates unreasonable clustering results. We also find that Fuzzy C-means possess monotonicity, which may generate meaningless clustering results. Aiming at these weak points, we present an improved Fuzzy C-means named IFCM (Improved Fuzzy C-means). Firstly, we research the reason of sensitivity and find that constraint leads to sensitivity of algorithm, we propose abolish constraint; secondly, we replace membership with typicality for acquiring more reasonable clustering results; finally, we add penalty factor to objective function to avoid monotonicity and coincident clustering results. On the basis of these, we improve objective function and provide step of algorithm. Experiments on various datasets show that new algorithm recognizes noisy data effectively and makes cluster effect improve furthermore.