摘要
针对区间不确定性数据的分类问题,提出一种基于超球支持向量机的多分类方法。采用超椭球凸集模型描述数据的不确定性信息;建立超球支持向量机的不确定约束规划模型,将其转化为两层嵌套约束规划问题;通过上下两层子优化交替迭代寻优的方法求解最优超球面,利用泰勒展开法,直接推导下层子优化线性近似问题的最优解,以降低计算复杂度。实验结果表明,该方法具有较高的分类精度及较好的抗噪性和鲁棒性,适合解决区间不确定性数据多分类问题。
Aiming at the problem of multi-classification for interval uncertainty data,a method based on hyper-sphere support vector machine（HSVM）was proposed.Firstly,uncertainty information of interval data was described using ellipsoidal convex model.Secondly,uncertainty constraint programming model of HSVM was established and converted to bi-level constraint programming problem.Finally,the optimization hyper-sphere of classification problem was obtained by alternating iterative optimization of the two sub optimization problems including upper and lower,and the approximate optimal solution of lower-sub optimization was directly derived using Taylor expansion.Experimental results show that this proposed method makes a better accuracy and has strong robustness for noise.It is effectively suitable for multi-class classification with data uncertainty.
出处
《计算机工程与设计》
北大核心
2015年第7期1778-1783,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(41362015)
江西省自然基金项目(20122BAB201045)
关键词
区间不确定性数据
超球支持向量机
超椭球凸集模型
非线性两层规划
分类
interval uncertainty data
hyper-sphere support vector machine
hyper-ellipsoid convex model
nonlinear bi-level programming
classification
