基于改进的粒子群算法和信息熵的知识获取方法  

Knowledge acquisition based on improved PSO algorithm and entropy

作　　者：许翔[1] 张东波[1] 黄辉先[1] 刘子文[1] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《计算机应用》2009年第8期2245-2249,共5页journal of Computer Applications

基　　金：湖南省自然科学基金资助项目(06JJ50112)

摘　　要：针对粒子群优化算法(PSO)易陷入局部优化的问题,在PSO算法加入交叉变异算子,克服了标准PSO算法易陷入局部最优的不足;并将改进的PSO算法和模糊C-均值聚类相结合,提出了一种新的模糊聚类算法CMPSO-FCM,该算法具有良好的搜索能力和聚类效果。进而将聚类得到的属性隶属矩阵用于属性约简,并提出一种基于信息熵的模糊粗糙集知识获取的方法。实验和实例分析表明该方法的正确性和有效性。Considering the problem that PSO algorithm is easy to fall into local optimum, crossover and mutation operators were introduced. The modified PSO algorithm was combined with Fuzzy C-Means （FCM） algorithm and a new fuzzy clustering algorithm （CMPSO-FCM） was proposed. The searching capability and clustering effect were improved by this new algorithm. Then the membership matrix obtained by clustering algorithm was used to reduce attribute set. Finally, based on entropy, a knowledge acquisition method of fuzzy Rough Set （RS） was put forward. Experiment and example were provided to verify the effectiveness and practicability of this approach.

关 键 词：粒子群优化 模糊C-均值 模糊粗糙集 属性约简 信息熵 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]