基于支持向量机的非对称区间回归分析

针对区间回归中上、下2个端点的误差范围不相同的非对称问题,建立了Fitness、Possibility 和Necessity 3个回归模型,对区间样本的中心趋势和最大、最小可能边界进行综合分析,并引入支持向量机,区分线性和非线性两种情况,提出了非对称区间回归支持向量机AIR-SVM(asymmetrical interval regression SVM)算法,对非对称区间数据集回归估计进行了分析.通过3个数据仿真实验,检验了提出算法的良好性能,有效地解决了非对称情况下精确数输入-区间数输出的区间数据回归问题.

Asymmetrical interval regression analysis based on SVM

The appro ach w as proposed for the asymmet rical interval regression problem, w hich meant theerrors affect ing tw o interval ends differ ing f rom each other. T hree models w er e established, w hich werefitness, possibility and necessity , to analyze the cent ral tendency , maximum and minimum bounds simultaneously. Fo r the linear and nonlinear regression condit ions, the estimat ion of asymmetr ical interval dataw as obtained by AIR-SVM( Asymmetr ical Interval Reg ressio n SVM, AIR-SVM) . The feasibil ity and goo dperformance of the alg orithm were show n by three no nlinear ex periments. T he asymmet rical problem ofinterval reg ressio n in crisp-input and interv al-output data sample regr ession est imat ion w as resolv ed suf ficiently .