论文部分内容阅读
FCM和PCM的混合模型可以克服它们单独聚类时的缺点,在聚类效果上有很大改进,但是对于特征不明显的样本而言,这种混合模型的聚类效果并不太好,为了克服这一缺点,本文引入Mercer核,提出了一种新的基于核的混合c-均值聚类模型(KIPCM),运用核函数使得在原始空间不可分的数据点在核空间变得可分。通过数值实验,得到了较为合理的中心值以及较高的正确分类率,证实了本文算法的可行性和有效性。
The hybrid model of FCM and PCM can overcome the shortcomings when they are clustered separately, and has great improvement on the clustering effect. However, the clustering effect of this hybrid model is not very good for the samples whose characteristic is not obvious, To overcome this shortcoming, Mercer kernel is introduced in this paper. A new kernel-based hybrid c-means clustering model (KIPCM) is proposed. By using kernel function, kernel data can make the data points inseparable in the original space separable in kernel space. Through numerical experiments, a more reasonable center value and a higher correct classification rate are obtained, which proves the feasibility and effectiveness of the proposed algorithm.