摘要
为提高极限学习机在失衡数据中的整体分类性能,提出一种基于代价敏感学习的自适应加权极限学习机分类算法。考虑各类间样本的差异性和同一类内样本的丰富性,利用类样本数量差异构造初始惩罚权重,分析样本附近异类样本数量确定额外代价权重,将两种代价权重相加构建自适应代价敏感惩罚矩阵。在公共数据集上的一系列对比实验结果表明,采用的自适应加权策略兼顾了不同类别样本的分布,在不平衡数据集上有效提高了算法整体分类精度。
To improve the overall classification performance of the extreme learning machine(ELM)on imbalanced data,an adaptive weighted extreme learning machine classification algorithm based on cost-sensitive learning was proposed.Both the varia-tions of samples between classes and the diversity of samples within the same class were considered.The difference in the number of samples from each class was applied to construct initial penalty weights,additional cost weights were determined by the number of heterogeneous samples near the example being analyzed.The sum of the two cost weights was employed to construct an adaptive cost-sensitive penalty matrix.The results of a series of comparison experiments on a public dataset show that the adaptive weighting strategy that takes into account the distribution of samples from different classes is employed and the overall classification accuracy of the algorithm is effectively improved on imbalanced dataset.
作者
孙中强
应文豪
毕安琪
王骏
龚声蓉
SUN Zhong-qiang;YING Wen-hao+;BI An-qi;WANG Jun;GONG Sheng-rong(School of Computer Science and Technology,Soochow University,Suzhou 215000,China;School of Computer Science and Engineering,Changshu Institute of Technology,Suzhou 215500,China;School of Communication and Information Engineering,Shanghai University,Shanghai 200444,China)
出处
《计算机工程与设计》
北大核心
2023年第7期2008-2014,共7页
Computer Engineering and Design
基金
国家重点研发计划基金项目(2018YFB1004901)
教育部人文社科基金项目(18YJCZH229)
江苏省教育科学十三五规划基金项目(X-a/2018/10)。
关键词
不平衡数据
加权极限学习机
代价敏感学习
自适应
分类
惩罚矩阵
类分布
imbalanced data
weighted extreme learning machine
cost-sensitive learning
adaptive
classification
penalty matrix
class distribution
