The main objectives of the study are to understand and explore critical feature wavelengths of the obtained near-infrared (NIR) data relating to dianhong black tea quality categories, we propose a multi-variable selection strategy based on the variable space optimization from big to small which is the kernel idea of a variable combination of the improved genetic algorithm (IGA) and particle swarm optimization (PSO) in this study. A rapid description based on the NIR technology is implemented to assess black tea tenderness and rankings. First, 700 standard samples from dianhong black tea of seven quality classes are scanned using a NIR system. The raw spectra acquired are preprocessed by Savitzky-Golay (SG) filtering coupled with standard normal variate transformation (SNV). Then, the multi-variable selection algorithm (IGA-PSO) is applied to compare with the single method (the IGA and PSO) and search the optimal characteristic wavelengths. Finally, the identification models are developed using a decision tree (DT), partial least-squares discriminant analysis (PLS-DA), and support vector machine (SVM) based on different kernel functions combined with the effective features from the above variables screening paths for the discrimination of black tea quality. The results show that the IGA-PSO-SVM model with a radial basis function achieves the best predictive results with the correct discriminant rate (CDR) of 95.28% based on selected four characteristic variables in the prediction process. The overall results demonstrate that NIR combined with a multi-variable selection method can constitute a potential tool to understand the most important features involved in the evaluation of dianhong black tea quality helping the instrument manufacturers to achieve the development of low-cost and handheld NIR sensors.

本研究的主要目的是了解和探索与滇红红茶质量类别有关的近红外（NIR）数据的关键特征波长，我们基于从大到小的可变空间优化，提出了一种多变量选择策略。这是本研究中改进遗传算法（IGA）和粒子群优化（PSO）的变量组合的核心思想。基于NIR技术的快速描述用于评估红茶的嫩度和等级。首先，使用近红外系统对七种质量等级的滇红红茶中的700个标准样品进行扫描。采集的原始光谱通过Savitzky-Golay（SG）滤波与标准正态变量变换（SNV）进行预处理。然后，应用多变量选择算法（IGA-PSO）与单一方法（IGA和PSO）进行比较，并搜索最佳特征波长。最后，使用决策树（DT），偏最小二乘判别分析（PLS-DA）和支持向量机（SVM），基于不同的核函数以及上述变量筛选路径的有效特征，开发了识别模型。用于区分红茶质量。结果表明，基于径向基函数的IGA-PSO-SVM模型基于预测过程中选择的四个特征变量，以95.28％的正确判别率（CDR）获得了最佳的预测结果。