| 摘要: |
| 提出了一种快速的稀有类检测算法——CATION(rare category detection algorithm based on weightedboundary degree).通过使用加权边界度(weighted boundary degree,简称WBD)这一新的稀有类检测标准,该算法可利用反向k 近邻的特性来寻找稀有类的边界点,并选取加权边界度最高的边界点询问其类别标签.实验结果表明,与现有方法相比,该算法避免了现有方法的局限性,大幅度地提高了发现数据集中各个类的效率,并有效地缩短了算法运行所需要的运行时间. |
| 关键词: 稀有类检测 边界点检测 加权边界度 k 近邻 反向k 近邻 |
| DOI：10.3724/SP.J.1001.2012.04104 |
| 分类号: |
| 基金项目:教育部-英特尔信息技术专项科研基金(MOE-INTEL-11-06) |
|
| Rare Category Detection Algorithm Based on Weighted Boundary Degree |
|
HUANG Hao, HE Qin-Ming, CHEN Qi, QIAN Feng, HE Jiang-Feng, MA Lian-Hang
|
|
College of Computer Science and Technology, Zhejiang University, Hangzhou 310027, China
|
| Abstract: |
| This paper proposes an efficient algorithm named CATION (rare category detection algorithm based on weighted boundary degree) for rare category detection. By employing a rare-category criterion known as weighted boundary degree (WBD), this algorithm can make use of reverse k-nearest neighbors to help find the boundary points of rare categories and selects the boundary points with maximum WBDs for labeling. Extensive experimental results demonstrate that this algorithm avoids the limitations of existing approaches, has a significantly better efficiency on discovering new categories in data sets, and effectively reduces runtime, compared against the existing approaches. |
| Key words: rare category detection boundary point detection weighted boundary degree k-nearest neighbor reverse k-nearest neighbor |
