用于特征子集选择的异步并行微粒群优化方法  

Asynchronous parallel particle swarm optimizer for feature subset selection

作　　者：孔莉芳[1,2] 张虹[1] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221116 [2]徐州空军学院基础部,江苏徐州221002

出　　处：《控制与决策》2012年第7期967-974,共8页Control and Decision

基　　金：国家自然科学基金项目(61005089);高等学校博士学科点专项科研基金项目(20100095120016);江苏省自然科学基金项目(BK2011215)

摘　　要：针对大量无关或冗余的特征通常会降低模式分类中分类器性能的问题,提出一种基于异步并行微粒群优化的特征子集选择方法(AP-PSO).该方法采用二进制微粒群优化搜索特征子集,利用异步并行方式提高算法的运算效率;为有效协调种群的全局探索和局部开发能力,充分利用混沌运动的遍历性和随机性,提出一种一致混沌变异算子.与已知4种特征子集选择方法进行比较,所得结果验证了该算法的有效性.In pattern classification system, many irrelevant and redundant features will lessen the performance of classifiers. Therefore, a feature subset selection method based on asynchronous parallel particle swarm optimization algorithm is proposed. This algorithm uses the binary particle swarm optimization to select feature subsets, and takes advantage of asynchronous parallel strategy to enhance time efficiency. In order to balance effectually the global exploration and the local exploitation of swarm, an uniform chaos mutation also is proposed by making the best use of the ergodicity, stochastic property and regularity of chaos. By compared with four known feature selection methods, the results show the effectiveness of the proposed algorithm.

关 键 词：特征子集选择 微粒群优化 混沌变异 异步并行 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]