Institute of Basic and Applied Chemistry of the Northeast of Argentina (IQUIBA-NEA), National Scientific and Technical Research Council (CONICET), Faculty of Exact and Natural Science and Surveying National University of the Northeast - UNNE, Av. Libertad 5470, 3400 Corrientes, Argentina; Chemistry Institute of Araraquara, São Paulo State University - UNESP, R. Prof. Francisco Degni 55, 14800-900 Araraquara, SP, Brazil.
Faculty of Agricultural Sciences, UNNE, Sgto. Cabral, 1213, 3400 Corrientes, Argentina.
Food Chem. 2020 Nov 30;331:127051. doi: 10.1016/j.foodchem.2020.127051. Epub 2020 Jun 15.
A simple, fast, and efficient spark discharge-laser-induced breakdown spectroscopy (SD-LIBS) method was developed for determining rice botanic origin using predictive modeling based on support vector machine (SVM). Seventy-two samples from four rice varieties (Guri, Irga 424, Puitá, and Taim) were analyzed by SD-LIBS. Spectral lines of C, Ca, Fe, Mg, N and Na were selected as input variables for prediction model fitting. The SVM algorithm parameters were optimized using a central composite design (CCD) to find the better classification performance. The optimum model for discriminating rice samples according to their botanical variety was obtained using C = 5.25 and γ = 0.119. This model achieved 96.4% of correct predictions in test samples and showed sensitivities and specificities per class within the range of 92-100%. The developed method is robust and eco-friendly for rice botanic identification since its prediction results are consistent and reproducible and its application does not generate chemical waste.
一种简单、快速、高效的火花放电激光诱导击穿光谱（SD-LIBS）方法，结合支持向量机（SVM）预测模型，用于鉴定稻米植物来源。通过 SD-LIBS 分析了四个稻米品种（Guri、Irga 424、Puitá 和 Taim）的 72 个样本。选择 C、Ca、Fe、Mg、N 和 Na 的光谱线作为预测模型拟合的输入变量。使用中心复合设计（CCD）优化 SVM 算法参数，以找到更好的分类性能。使用 C=5.25 和γ=0.119 获得了根据植物品种区分稻米样品的最佳模型。该模型在测试样本中达到了 96.4%的正确预测率，每个类别均表现出 92-100%的灵敏度和特异性。由于其预测结果一致且可重复，并且其应用不会产生化学废物，因此该方法对于稻米植物鉴定具有强大且环保的优势。
