摘要
从计算方法角度对算法进行改进,引入高斯核函数,改良归一化条件并对迭代过程加以简化,从而改进了模糊核C均值算法。算法性能速度较经典的聚类算法有了较大改进,聚类结果更为快速稳定,并可在多种数据结构条件下进行有效的聚类,计算时间显著减少,克服了传统的模糊核C-均值算法计算时间较长,在样本集不理想的情况下可能导致结果不好等不足。实验结果证实了该改进算法有效性。
This paper presents a modified fuzzy clustering algorithm based on the Gauss function kernel method （ MFKCM ）, which is more effective than the classical fuzzy kernel C means clustering algorithm （FKCM） in many aspects, such as shorter computing times and more accurate result. The modified algorithm integrates the Gauss kernel function and improved FCM algorithm, has better robustness, and gets better clustering effect in the circumstance of outliers existing. Experimental results show that MFKCM can effectively cluster data with diversiform structures in contrast to FCM and FKCM.
出处
《电子科技》
2008年第10期49-51,55,共4页
Electronic Science and Technology
基金
国家自然科学基金资助项目(60672026)
关键词
模糊C-均值算法
核函数
高斯核函数
模糊核C-均值算法
fuzzy C-means algorithm
kernel function
gauss function
fuzzy kernel C-means clustering algorithm
