气相色谱结合化学计量在核桃油、菜籽油掺伪混合物含量判别分析中的应用

目的：识别核桃油中同时掺入多种其他植物油的多元掺伪。方法：采用气相色谱技术分析核桃油、菜籽油掺伪混合物中脂肪酸含量，结合化学计量方法对气相色谱数据进行建模，并对不同比例核桃油、菜籽油混合物进行判别分析。结果：主成分分析法（PCA）能快速鉴别出纯核桃油和掺伪核桃油，并在一定程度上反映样本的掺伪比例；Bayes判别分析将83.33%的样品归入正确的分类；偏最小二乘判别分析法（PLS-DA）的判别准确率达87.50%；基于BP神经网络模型的判别分析，其训练集的正确率为84.21%，测试集的正确率为80.00%；基于遗传算法优化支持向量机（SVM-ga）的判别分析，其训练集和测试集的正确率均为100%。结论：多种分析模型均能不同程度地识别核桃油、菜籽油掺伪比例，其中SVM-ga模型的预测精度最佳。

Application of gas chromatography combined with stoichiometry in discriminant analysis of pseudo mixtures of walnut oil and rapeseed oil

Objective: This study aimed to investigate different discriminant analysis models for pseudo-mixtures of walnut oil and rapeseed oil. Methods: Gas chromatography technology was used to analyze the fatty acid content in the adulterated mixture of walnut oil and rapeseed oil. Chemical stoichiometric methods were used to model the gas chromatography data, and discriminant analysis was performed on different proportions of walnut oil and rapeseed oil mixtures. Results: Pure walnut oil and adulterated walnut oil were distinguihed by using Principal Component Analysis (PCA) identified, and percentage of adulteration in the sample was calculated. 83.33% of the samples were successfully categorized using the Bayes discriminant analysis. Partial Least Squares Discriminant Analysis (PLS-DA) achieved 87.50% discrimination accuracy. Based on the BP neural network model for discriminant analysis, the accuracy of the training set was 84.21% and the accuracy of the test set was 80.00%. For both the training and testing sets, the genetic algorithm-based discriminant analysis using an optimized support vector machine (SVM-ga) achieved 100% accuracy. Conclusion: Multiple analytical models can identify the adulteration ratio of walnut oil and rapeseed oil to varying degrees, among which the SVM-ga model had the best prediction accuracy.