基于假设间隔的弱随机特征子空间生成算法

Weak Random subspace Based On Simba

集成算法是机器学习领域的研究热点。随机子空间算法是集成算法的一个主要算法。随机子空间生成的特征子集可能含有冗余特征、噪声特征,影响算法的分类精度。为此,本文提出了一种基于假设间隔的弱随机特征子空间生成算法(WRSSimba),有效去除了特征子集中冗余特征和噪声特征。在UCI数据集上的实验结果表明,WRSSimba的分类性能优于随机子空间算法和Simba算法。

The ensemble algorithm is a hot research field of machine learning. Random subspace algorithm is a main algorithm of ensemble algorithm. Feature subset generated by random subspace may contain redundant feature and even noise feature, affecting the classification accuracy. Therefore, in this paper, Weak Random subspace Based On Simba （WRSSimba） algorithm is introduced. WRSSimba effectively eliminates the redundancy and noise feature of feature subspace. The experimental results on UCI datasets show that, WRSSimba classification performance is better than Random subspace algorithm.