Abstract: To improve the detection effect in oil adulteration, a new algorithm called ML-LVQ (Multi-Label Learning Vector Quantization) was proposed, which adapted Learning Vector Quantization (LVQ) to solve the multi-label learning problem on High Performance Liquid Chromatography (HPLC) data. It could minimize the upper bound of the ranking error, which would benefit the ranking measure. Moreover, the meta-labeler was used to identify the number of the labels for improving the bipartitions measure. The experimental results on nine classes of pure oil and their mixed oil samples show that the proposed algorithm is superior to the improved AdaBoost.RMH.

Key words: multi-label algorithm, Learning Vector Quantization (LVQ) algorithm, meta-labeler, High Performance Liquid Chromatography (HPLC) method, oil adulteration detection

摘要： 为了提高食用油掺伪检测效果，基于食用油的高效液相色谱数据，提出了一个新的多标号学习矢量量化算法(ML-LVQ)，并应用于食用油的掺伪检测中。它每次调整两个原型使排序损失的上界最小，并通过元标号分类器确定多标号的数目，从而达到同时优化ranking准则函数和bipartitions准则函数的目的。在9类纯油以及它们的混合油样本的数据集上测试的结果表明，ML-LVQ取得了比改进的AdaBoost.RMH算法更好的性能。

关键词: 多标号算法, 学习矢量量化算法, 元标号分类器, 高效液相色谱法, 食用油掺伪检测