类不平衡模糊加权极限学习机算法研究

发布时间：2018-03-23 02:26

本文选题：极限学习机　切入点：类不平衡学习　出处：《计算机科学与探索》2017年04期 　论文类型：期刊论文

【摘要】：从理论上分析了样例不平衡分布对极限学习机性能产生危害的原因;在该理论框架下探讨了加权极限学习机在处理此类问题上的有效性及其固有缺陷;引入模糊集的思想,对传统的加权极限学习机进行了改进,并提出了4种用于解决类不平衡问题的模糊加权极限学习机算法;最后通过20个基准的二类不平衡数据集对所提算法的有效性和可行性进行了验证。实验结果表明:较之加权极限学习机及几种传统的不平衡极限学习机算法,提出的算法可明显获得更优的分类性能,并且与模糊加权支持向量机系列算法相比,所提算法通常可获得与之相当的分类性能,但时间开销往往更小。
[Abstract]:In this paper, the causes of damage to the performance of extreme learning machine caused by sample unbalance distribution are analyzed theoretically, the validity and inherent defects of weighted ultimate learning machine in dealing with such problems are discussed under the framework of this theory, and the idea of fuzzy set is introduced. The traditional weighted ultimate learning machine is improved, and four fuzzy weighted ultimate learning machine algorithms are proposed to solve the class imbalance problem. Finally, the validity and feasibility of the proposed algorithm are verified by two kinds of unbalanced data sets of 20 benchmarks. The experimental results show that compared with the weighted extreme learning machine and several traditional unbalanced ultimate learning machine algorithms, Compared with the fuzzy weighted support vector machine series algorithm, the proposed algorithm can obtain the same classification performance, but the time cost is often less than that of the fuzzy weighted support vector machine algorithm.
【作者单位】：江苏科技大学计算机科学与工程学院;东南大学自动化学院;
【基金】：国家自然科学基金Nos.61305058,61471182,61572242 江苏省自然科学基金Nos.BK20130471,BK20150470 中国博士后科学基金Nos.2013M540404,2015T80481 江苏省博士后基金No.1401037B~~
【分类号】：TP181

【相似文献】