基于加权样本的FCM快速算法研究

A Fast FCM Algorithm Based on Weighted Sample

为改进FCM算法在处理大样本集聚类时速度慢、耗时多的缺点,根据样本在特征空间中的特征值分布情况,引入等价样本和样本加权概念,在此基础上提出了FCM(Fuzzy C-Means)的快速算法一般形式:WFCM(WeightedFuzzy C-Means)算法。理论上证明了WFCM算法和FCM算法对样本集分割的等价性,并且,WFCM在运算性能方面明显优于FCM算法。而两个算法在灰度图像分割上的例子验证了WFCM算法的快速性和有效性。

To improve the computational performance of the FCM algorithm used in the dataset clustering with large numbets, the concepts of the equivalent samples and the weighting samples based on the samples＇ eigenvalue distribution in the feature space were introduced and a novel fast cluster algorithm named WFCM （weighted fuzzy C-means） algorithm was put forward, which was inherited from the traditional FCM algorithm. It was proved that the cluster results were equivalent in dataset with two different cluster algorithms： WFCM and FCM. Otherwise, the WFCM algorithm had better computational performance than the ordinary FCM algorithm. The experiment of the gray image segmentation showed that the WFCM algorithm is a fast and effective cluster algorithm.