中文摘要：

针对复杂产品客户需求中类别数多且数据量大的分类问题,利用最小最大概率机的概率信息和样本间隔信息,提出采用启发式方法解决编码设计问题。在算法中将一个减少了迭代次数的分类器视为弱化了的分类器,同时保留分类器的间隔属性和几何特性,并利用高维映射将分类器输出映射到特征空间,在弱化阶段大量减少了整体的迭代次数而在合成阶段利用非线性映射来提升性能。通过对叉车产品客户需求的数值试验证明算法具有较好的分类效果,降低了对二类分类器的精度要求。

英文摘要：

Aiming at the classification problem of the complex product voice of customers with a large number of catalogs and data volumes,a heuristic strategy was proposed to resolve the coding designing problem by using output of a probability minimax machine and sample interval attribute.An algorithm with less iterative steps of optimization was taken as a ＂weak＂ algorithm while preserving its other characteristics like large margin and geometric properties.The nonlinear mapping was applied for ＂weak＂ algorithms＇ output（vector） to work in high dimensional spaces and improve the performances.At last,the experiments with the proposed algorithm were described to show that the approach was more effective than others by using forklift VOC.