摘要
针对二进制粒子群算法(BPSO)具有过早收敛的缺陷,在粒子位置更新后提出变异概率自适应从大到小的变异操作。同时对算法惯性权重参数采用递增的设置方案,从而得到一种自适应变异BPSO算法(AMBPSO),将其应用于特征选择问题。实验结果表明,提出的新算法前期具有较强的全局搜索能力,后期具有较强的局部搜索能力,能使平均选择特征数量最多从27.6个减少到20.2个,平均分类准确率最多从91.346%提升到94.135%。
Aiming at the defect of premature convergence of the binary particle swarm algorithm(BPSO),a mutation operation with the adaptation of the mutation probability going from large to small was proposed after particle position updating.An incremental setting scheme was adopted for the inertia weight parameters of the algorithm to obtain an adaptive mutation BPSO algorithm(AMBPSO),which was applied to the feature selection problem.Experimental results show the proposed new algorithm has strong global search ability in the early stage,and has strong local search ability in the later stage,and it can reduce the average number of selected features from 27.6 to 20.2 and increase the average classification accuracy from 91.346%to 94.135%.
作者
姜磊
刘建华
张冬阳
卜冠南
JIANG Lei;LIU Jianhua;ZHANG Dongyang;BU Guannan(School of Information Science and Engineering, Fujian University of Technology, Fuzhou 350118, China;Fujian Provincial Key Laboratory of Big Data Mining and Applications, Fuzhou 350118, China)
出处
《福建工程学院学报》
CAS
2020年第3期273-279,共7页
Journal of Fujian University of Technology
基金
福建省自然科学基金项目(2019J01061137)
福建工程学院发展基金(GY-Z17150)。
