基于SVM的精确数-区间数回归模型建模方法

SVM Based Algorithm for Regressive Modeling with Accurate Data Input-interval Number Output

分析了现有的精确数输入和区间数输出回归算法存在的问题,提出了基于支持向量机的区间数回归建模算法.该算法把支持向量机从精确数回归分析方法推广到区间数回归分析建模方法,在小样本训练集下回归模型具有良好的泛化性能,有效地避免了现有算法中回归模型的下界可能大于上界的问题.以连续退火生产过程中冷却段出口带钢温度预测为例,通过仿真说明了该算法的有效性.

A SVM based algorithm is proposed to cope with regressive modeling problems with accurate data inputinterval number output. A regressive modeling approach based on the support vector machine （SVM） theory is generalized from real number domain to interval number domain. The proposed algorithm can promote the generalization properties of regressive models in the ease of small sampling sets and effectively overcomes the shortage of the existing algorithms that the lower limits of model outputs may exceed their upper limits. An example for the prediction of temperature distribution of export strap steel in continuous annealing process demonstrates the effectiveness and efficiency of the proposed method.