摘要
针对非线性多分类问题,提出了一个改进的嵌入最小-最大值特征选择算法,并与支持向量机算法结合,提出了针对复杂的组合优化问题的启发式算法。为验证方法的有效性,在钢板缺陷识别工程数据集上进行了实验,表明所提出的方法具有较高的求解速度和预测准确度。
An improved embedded min-max feature selection algorithm was proposed for the nonlinear multi-label classification problem,and in combination with the support vector machine algorithm,a heuristic algorithm was proposed for the complex combinatorial optimization problem.The efficiency and accuracy of the proposed algorithm were verified after a series of experiments conducted on steel faults diagnosis dataset.
作者
武小军
周文心
董永新
WU Xiaojun;ZHOU Wenxin;DONG Yongxin(School of Economics and Management,Tongji University,Shanghai 200092,China)
出处
《同济大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2022年第2期153-159,共7页
Journal of Tongji University:Natural Science
关键词
最小-最大值优化问题
特征选择
非线性多分类支持向量机
min-max optimization
feature selection algorithm
nonlinear multi-label support vector machine