The volatile compounds of super fresh, superior fresh, general fresh, corrupt and completely corrupt fish meal were studied by headspace solid-phase microextraction–gas chromatography–mass spectrometry. Principal component analysis was used to analyze the sensor and gas chromatography-mass spectrometry data, so as to study the resolution of sensor array. A total of 198 fish meal samples with different freshness were classified by the gradient boosting decision tree method, yielding a model between the sensor array data and the freshness index, such as the acid value and the volatile base nitrogen value. The gradient boosting decision tree model shows that the correlation between the predicted acid value by electronic nose and the measured acid value is 0.90 and the correlation between the predicted volatile base nitrogen value and the measured volatile base nitrogen is 0.97. The combination of an electronic nose and a pattern recognition method based on a gas sensor can predict the acid value and volatile base nitrogen of the freshness index and can detect the freshness of fish meal.

中文翻译：

---

GBDT基于便携式电子鼻系统和HS-SPME–GC–MS的鱼粉新鲜度检测

通过顶空固相微萃取-气相色谱-质谱法研究了超新鲜，优质新鲜，普通新鲜，腐败和完全腐败的鱼粉中的挥发性化合物。利用主成分分析法对传感器和气相色谱-质谱数据进行分析，以研究传感器阵列的分辨率。通过梯度增强决策树方法对总共198个不同新鲜度的鱼粉样品进行分类，从而在传感器阵列数据和新鲜度指数（例如酸值和挥发性碱氮值）之间建立模型。梯度提升决策树模型表明，电子鼻的预测酸值与测得的酸值之间的相关性为0。在图90中，预测的挥发性基础氮值与测量的挥发性基础氮之间的相关性是0.97。电子鼻和基于气体传感器的模式识别方法的结合可以预测新鲜度指数的酸值和挥发性碱氮，并可以检测鱼粉的新鲜度。