基于Boltzmann选择的改进ABFM算法

Modified ABFM algorithm based on Boltzmann selection

模糊C-均值(FCM)聚类算法是数据挖掘中应用广泛的一种方法,但还存在容易陷入局部极小值和对初始值敏感的缺点,为此提出了一种基于Boltzmann选择机制的改进人工蜂群的模糊C-均值聚类算法(BABFM)。该算法引入了Boltzmann选择机制代替轮盘赌的选择方式,采用小区间生成法使初始群体均匀化,使得该算法的全局寻优能力更强,有效克服了FCM算法的缺点。实验结果表明,新算法与FCM和ABFM聚类算法相比聚类效果更准确,效率更高,迭代次数更少。

Fuzzy C-Means（FCM） clustering algorithm is one of the most widely used methods in data mining, but there still exist some defects, such as the local optima and sensitivity to initialization.Therefore,a new modified FCM clustering algorithm（BABFM） is proposed based on ABFM.BABFM algorithm applies Boltzmann selection mechanism instead of roulette and uses min intervals to make the initial group more symmetrical and have better capacity of global search.It can effectively solve faults of FCM.According to the test, the new algorithm is more accurate in clustering, higher efficiency and fewer iterations than FCM and ABFM clustering algorithm.