基于改进离散二进制粒子群的SVM选择集成算法

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.