摘要
提出了基于优化的随机子空间分类集成算法CEORS,该算法通过运用封装式特征选择和LSA降维两种方法对随机选择的特征子集进行了优化,并运用优化的特征子空间进行分类器的集成.实验结果表明,基于优化特征子空间的集成分类器性能优于Bagging和AdaBoost.
In this paper, we propose a classifier ensemble algorithm based on optimal random subspace （CEORS）, this algorithm firstly uses wrapper feature selection and LSA technology to optimize the selected feature subspace, then constructs the basic classifier using the optimal feature subspaces, and at last gets ensemble classifier by integrate all the basic classifiers. Numerical experimental results show that the CEORS algorithm has better and comparable classification as compared to Bagging and AdaBoost.
出处
《微电子学与计算机》
CSCD
北大核心
2009年第10期158-160,164,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(60873176)
