摘要
在数据预处理过程中模糊c-均值对初值的设定非常敏感,如果初值设置不好容易陷入局部最优解。提出了用减法聚类对模糊c-均值进行初值设定的方法,并把其应用到了模糊规则的生成过程中,通过实验可知应用此种方法不但可以得到全局最优解,加快收敛速度而且不必事先给定聚类个数,所以这是一种行之有效的数据预处理方法。
It is very sensitivity to its initial value when we use fuzzy c-means（FCM）to put up data pretreatment. It will get into local optimum solution if the setting of initial value is not good. The thesis puts forward a ways that can use subtractive clustering to initialize the initial value of FCM and uses this means in the course of generating the fuzzy rules. The experiments show that this way can gain the optimum solution,quicken the rate of convergence and we need not to give the cluster number beforehand. So it is a good way of data pretreatment.
出处
《中国民航大学学报》
CAS
2007年第2期36-39,共4页
Journal of Civil Aviation University of China
基金
中国民航大学科研基金(06kym02
05yk15m)
