极限学习机类不平衡数据学习算法研究

尽管极限学习机因具有快速、简单、易实现及普适的逼近能力等特点被广泛应用于分类、回归及特征学习问题,但是,极限学习机同其他标准分类方法一样将最大化各类总分类性能作为算法的优化目标,因此,在实际应用中遇到数据样本分布不平衡时,算法对大类样本具有性能偏向性。针对极限学习机类不平衡学习问题的研究起步晚,算法少的问题,在介绍了极限学习机类不平衡数据学习研究现状,极限学习机类不平衡数据学习的典型算法-加权极限学习机及其改进算法的基础上,提出一种不需要对原始不平衡样本进行处理的Adaboost提升的加权极限学习机,通过在15个UCI不平衡数据集进行分析实验,实验结果表明提出的算法具有更好的分类性能。

Extreme learning machine for imbalance data learning

Extreme learning machine because of some characteristics such as adaptability, learning capability, generalization ability was applied as the estimator in regression problem or the classifier for classification tasks. Though the standard ELM owned better generalization performance compared with many other machine learning methods, ELM was not well-suited for imbalance dataset classification in its basic form, since it had a natural tendency to favor the majority class by assuming balanced class distribution or equal misclassification cost when classifying data with complex class distribution. In this paper, we reviewed the recent theoretical advances related to weighted extreme learning machine(WELM), introduced the extensions and improvements of WELM and proposed a weighted extreme learning machine based on Adaboost to obtain better performance than WELM. The experimental results on 15 imbalanced datasets showed that the proposed method indicates its superiority.