基于改进离散二进制粒子群的SVM选择集成算法  

SVM selection ensemble algorithm based on improved binary particle swarm optimization

作　　者：孟常亮[1] 李卫忠[1] 廖勇[1,2] 华继学[1] 

机构地区：[1]空军工程大学导弹学院 [2]中国人民解放军95824部队

出　　处：《计算机工程与应用》2011年第28期166-168,231,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60975026);陕西省自然科学基金(No.2007F19)~~

摘　　要：针对基于离散二进制粒子群(BPSO)的SVM选择集成算法的分类精度不高,以及所选分类器个数过多等问题,利用改进的离散二进制粒子群算法(IBPSO)和SVM选择集成算法相结合,提出基于IBPSO的SVM选择集成算法。通过选用合适的适应度函数以及调节因子k,进行多次仿真,实验表明,对由boostrap方式生成的SVM集合,基于IBPSO的SVM选择集成在精度和分类器个数方面均优于基于BPSO的SVM选择集成,证明了IBPSO算法的优越性。For the low classification precision and excessive classifiers of SVM selection ensemble algorithm based on improved binary particle swarm optimization, this paper combines IBPSO and SVM selection ensemble algorithm to bring forward SVM selection ensemble algorithm based on IBPSO.As suitable fitness function and regulatory factor k are selected, for the SVM sets are generated by boostrap,both the precision and the number of classifiers of SVM selection ensemble algorithm based on IBPSO are superior to those of selection ensemble algorithm based on BPSO.The experiment proves the superiority of the former algorithm.

关 键 词：离散二进制粒子群 支持向量机(SVM)选择集成 适应度函数 调节因子 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]