Computer Engineering and Applications ›› 2016, Vol. 52 ›› Issue (13): 121-125.

Previous Articles     Next Articles

Study on classifier ensemble learning model for extreme learning machine

SHAO Liangshan, MA Han, WEN Tingxin   

  1. School of Business Administration, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2016-07-01 Published:2016-07-15

极限学习机的分类器集成模型研究

邵良杉,马  寒,温廷新   

  1. 辽宁工程技术大学 系统工程研究所,辽宁 葫芦岛 125105

Abstract: A RF-ELM?classifier ensemble model is proposed, which both combines rotation forest algorithm and extreme learning machine algorithm. This model utilizes extreme learning machine to be base classifier and rotation forest algorithm to be integration framework. Three experiments are completed on eight data sets. According to the experimental results, the effects of hidden layer neurons number to forecast results and the instability defects of a single ELM prediction model are discussed. The results of contrast experiment for stability and prediction accuracy demonstrate that ensemble learning model for ELM algorithm can improve its performance and RF-ELM has a better stability and precision than ELM algorithm and Bagging-ELM model. RF-ELM model is proved to be an effective classifier ensemble learning model for ELM algorithm.

Key words: Extreme Learning Machine(ELM), Rotation Forest(RF), classifier ensemble, Bagging algorithm

摘要: 将极限学习机算法与旋转森林算法相结合,提出了以ELM算法为基分类器并以旋转森林算法为框架的RF-ELM集成学习模型。在8个数据集上进行了3组预测实验,根据实验结果讨论了ELM算法中隐含层神经元个数对预测结果的影响以及单个ELM模型预测结果不稳定的缺陷;将RF-ELM模型与单ELM模型和基于Bagging算法集成的ELM模型相比较,由稳定性和预测精度的两组对比实验的实验结果表明,对ELM的集成学习可以有效地提高ELM模型的性能,且RF-ELM模型较其他两个模型具有更好的稳定性和更高的准确率,验证了RF-ELM是一种有效的ELM集成学习模型。

关键词: 极限学习机, 旋转森林, 分类器集成, Bagging算法