| 摘 要: 传统的支持向量机分类算法在面对分类环境中的复杂噪声问题时往往存在局限性。针对这一问题提出一种基于聚类表征的支持向量机分类方法。利用样本之间的距离信息自适应学习具有一定概率邻域分配的结构化图。此外,通过邻域样本构造一个新的数据样本能够增强其数据表征,有助于更好地避免分类器受到噪声干扰。通过原数据下以及大量噪声诱导场景下的各项实验验证,实验结果平均提升了4%,证明该方法在解决分类噪声问题上具有显著的效果。 |
| 关键词: 支持向量机 样本空间 噪声 聚类 表征学习 |
|
中图分类号: TP181
文献标识码: A
|
| 基金项目: 国家自然科学基金项目(62376108) |
|
| Support Vector Machine Classification Method Based on Clustering Representation |
|
JIANG Lulu, SHEN Xiangjun
|
(School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, China)
2544998926@qq.com; xjshen@ujs.edu.cn
|
| Abstract: Traditional Support Vector Machine (SVM) classification algorithms often exhibit limitations when dealing with complex noise problems in classification environments. To address this issue, this paper proposes a clustering representation-based SVM classification method. It adaptively learns a structured graph with probabilistic neighborhood assignments by utilizing distance information between samples. Furthermore, constructing a new data sample using neighborhood samples enhances its data representation, which helps the classifier better avoid noise interference. Experimental results conducted on the original data and under numerous noise-induced scenarios demonstrate that the proposed method achieves an average improvement of 4 percentage points, proving its significant effectiveness in solving classification noise problems. |
| Keywords: support vector machine (SVM) sample space noise clustering representation learning |
