摘要
根据模糊 C-均值(FCM)算法中的类中距与类间距构造一个新且简单的分类准则函数vZS ,利用迭代自组织分析技术(ISODATA)和遗传算法(GA)嵌套构成遗传-迭代自组织分析技术(GA-ISODATA)共同执行 FCM 算法的优化计算.通过与同类方法比较,该方法不仅能够在给定预分类数的前提下实现最优分类,而且可以在完全不需要人工干预的环境下直接根据分类准则得到模糊 C-均值中的最优分类与相应的最佳分类数.当运用其他分类准则进行分类计算时只需要修改遗传算法中的适应度函数,所以 GA-ISODATA 具有很强的普适性.
We construct a new and simple classify rule based on intra-distance and inter-distance of fuzzy C-means (FCM) and nests ISODATA and GA to construct the GA-ISODATA in order to perform optimization computing of the FCM. Comparing with similar methods, this method can not only complete the optimal partition on the promise of giving the number of pre-classify, but also directly get the optimal number of classify in FCM without people-engaged. We only need to change the fitness function in GA when doing optimization computing for different classify rules, so GA-ISODATA also fits for optimal classify and the computing of optimal classify number of other rules.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2005年第3期52-61,共10页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(70273044)