电子学报 ›› 2018, Vol. 46 ›› Issue (11): 2714-2724.DOI: 10.3969/j.issn.0372-2112.2018.11.019

• 学术论文 • 上一篇    下一篇

基于模糊C-均值的相似性特征转换光滑支持向量机

方佳艳1,2, 刘峤1,2, 吴德3, 秦志光1,2   

  1. 1. 电子科技大学信息与软件工程学院, 四川成都 611731;
    2. 网络与数据安全省级重点实验室(电子科技大学), 四川成都 611731;
    3. 西安电子科技大学计算机学院, 陕西西安 710071
  • 收稿日期:2017-10-15 修回日期:2018-02-21 出版日期:2018-11-25 发布日期:2018-05-16
  • 通讯作者: 吴德
  • 作者简介:方佳艳 女,1997年出生,安徽池州人.电子科技大学信息与软件工程学院本科生,主要研究方向为机器学习、人工智能.E-mail:fyk80@163.com
  • 基金资助:
    国家863高技术研究发展计划(No.2011AA010706);国家自然科学基金重点项目(No.61133016,No.61772117);四川省高新技术及产业化面上项目(No.2017GZ0308)

Smooth Support Vector Machine with Similarity-Based Feature Transformation Technique and Fuzzy C-Means Clustering

FANG Jia-yan1,2, LIU Qiao1,2, WU De3, QIN Zhi-guang1,2   

  1. 1. School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu, Sichuan 611731, China;
    2. Provincial Key Laboratory of Network and Data Security, Chengdu, Sichuan 611731, China;
    3. School of Computer Sciences, Xidian University, Xi'an, Shaanxi 710071, China
  • Received:2017-10-15 Revised:2018-02-21 Online:2018-11-25 Published:2018-05-16

摘要: 在用于非线性分类的光滑支持向量机(SSVM)模型中,核函数必须满足Mercer's条件,由此限制了核函数的选择范围;并且在面对大规模数据集时,SSVM模型的计算复杂度很高,训练时间长.针对这两点缺陷提出了基于模糊C-均值的相似性特征转换光滑支持向量机模型(SFT-SSVM-FCM).首先,运用基于相似性的特征转换,使得核函数不需要再满足Mercer's条件,从而拓宽了核函数的选择范围;其次,运用模糊C-均值(FCM)分群技术,将完整的训练数据集划分成若干子簇,分别在每一个子簇上进行已经过相似性特征转换的SSVM模型训练.实验表明:与传统的SVM、SSVM模型及一系列变体模型相比较,该新模型在训练时间、分类精度方面都具有更好的表现.

关键词: 光滑支持向量机, 模糊C-均值, 相似性, 特征转换

Abstract: We propose a new model called smooth support vector machine with similarity-based feature transformation and fuzzy C-means (FCM) clustering (SFT-SSVM-FCM).When the similarity-based feature transformation technique is applied,Mercer's conditions are no longer required for kernel functions,thus broadening the range of usable kernel functions.We also incorporate the Fuzzy-C means clustering technique to divide a whole dataset into several clusters each of which is used to perform SSVM with similarity-based feature transformation.The experimental results indicate that the proposed model has better performance compared with the conventional SVM and SSVM model as well as some variants in terms of classification accuracy and training time.

Key words: smooth support vector machine, fuzzy C-means (FCM), similarity, feature transformation

中图分类号: