重庆理工大学学报2025，Vol.39Issue(11)：108-116,9.DOI:10.3969/j.issn.1674-8425(z).2025.06.013
面向非平衡数据集的深度极限学习机模型
Deep extreme learning machine model for imbalanced datasets
摘要
Abstract
To address the adverse effects of imbalanced datasets on the performance of classifiers,this paper proposes an adaptive deep extreme learning machine model based on Universum and improved Harris Hawk Optimization algorithm.First,the data preprocessing stage introduces Universum samples to enhance the learning of different classes of samples.Then,an improved Harris Hawk Optimization algorithm is proposed.The improved algorithm redefines the position update formulas to boost search performance and optimizes the minimization problem to increase stability.These modifications yield more precise parameters for optimizing weight parameters in the Class-specific Cost Regulation Extreme Learning Machine.Finally,based on Multi-layer Perceptron theory,an adaptive is built to determine optimal hidden layer parameters,further enhancing classification performance.The experimental results based on public datasets show the proposed adaptive deep extreme learning machine model markedly improves the classification performance of minority samples.Further predictions of classification results using a stroke screening dataset show the model provides some auxiliary diagnosis suggestions for clinical data.关键词
非平衡数据集/极限学习机/哈里斯鹰算法/UniversumKey words
imbalanced datasets/extreme learning machine/Harris Hawk optimization/Universum分类
信息技术与安全科学引用本文复制引用
张喻喻,李凤莲,王伟丽,贾文辉,黄丽霞,陈桂军..面向非平衡数据集的深度极限学习机模型[J].重庆理工大学学报,2025,39(11):108-116,9.基金项目
国家自然科学基金项目(62171307) （62171307）
