摘要
设计了一种针对不平衡数据集的学习机,即组合降采样极限学习机(EUS ELMs)。当训练数据集不平衡时,普通分类器对少数样本的分类敏感性较低,而给予多数样本过度关注。针对这种问题,将组合降采样结构与极限学习机(ELM)结合起来,形成一种新的组合学习机。该学习机继承了组合降采样结构对样本选择的依赖性小的优点,和极限学习机分类效率高、耗时短的优势,而且可以通过不同的参数设置控制对少数样本的关注度,根据使用者需求获得不同分类效果。
This paper presents the design of a learning machine targeting data imbalance. When the number of patterns from a class is much larger than that from the other class, standard classifiers are likely to be overwhelmed by the large class while ignoring the small one. It often degrades the classification performance. This paper, to address this problem, proposes an ensemble of under-sampled ELMs or EUS ELMs. The ensemble scheme can lower the variation of sample selection and the use of ELM can enhance the classification efficiency significantly. In addition, by adjusting the parameters of the learning machine, the user can alter the emphasis on different classes.
出处
《信息技术》
2015年第11期159-162,共4页
Information Technology
