论文部分内容阅读
针对现有支持向量数据描述(SVDD)快速决策方法在检测不同分布特性的未知样本时分类精度低下的问题,提出基于超椭球分类面的SVDD(HE-SVDD)快速决策方法.该方法通过构建超椭球分类面,提高了不同分布类型数据的分类精度,同时将SVDD的决策复杂度从O(n)降低到O(2)(n为支持向量数量).首先研究超球分类面快速决策方法的局限性,进而给出超椭球分类面的构建方法.在多种数据集上的实验结果表明,HE-SVDD可以在很大程度上提升现有快速决策方法的分类精度和适用数据类型.
Aiming at the problem of low accuracy of SVDD method when detecting unknown samples with different distribution characteristics, a fast hyperspheric classification-based SVDD (HE-SVDD) decision-making method is proposed in this paper. Hyperspheric classification surface to improve the classification accuracy of different distribution type data, and reduce the complexity of SVDD decision from O (n) to O (2) (n is the number of support vectors) .Firstly, The limitations of the method and the construction method of the hyperellipsoid classification surface.Experimental results on various datasets show that HE-SVDD can greatly improve the classification accuracy and applicable data types of the existing rapid decision methods .