Published in Scientific Papers. Series D. Animal Science, Vol. LXVIII, Issue 2
Written by Ruiqin MA, Xinxing LI, Norian POPOVICI, Gabriela VARIA, Carmen Georgeta NICOLAE, Liliana Mihaela MOGA
Mutton is an important part of the national diet, and its quality and safety are directly related to people's health. Therefore, accurate non-destructive testing of mutton freshness can effectively improve the level of food safety and ensure public health. In this paper, fresh lake mutton was taken as the research object, starting from the quality change mechanism of mutton, and the near-infrared spectrum data of fresh mutton was collected by using near-infrared spectrometer. In order to improve the computational efficiency and model performance, a principal component analysis model was established to integrate the NIR spectral data and extract the most important nine principal component features. Then, five near-infrared spectral classification models including logistic regression, SVM, SGD, decision tree and random forest were established to predict volatile basic nitrogen (TVB-N) to detect the freshness of mutton. The freshness prediction of decision tree model and random forest model has high overall prediction accuracy, which is more than 95%. The recognition rates of "Level 1" and "Level 2" samples in the decision tree model are 94.44% and 96.78%, respectively. The recognition rates of the random forest model for the "Level 1" and "Level 2" samples of the prediction set are 100% and 96.88%, respectively. Compared with the decision tree model, the random forest model improves the recognition rate of "Level 1" and "Level 2" by 5.56% and 0.1% respectively. The random forest model shows excellent prediction performance as the best classification model, with an accuracy of 97.96%. In this study, the potential of using NIR spectroscopy and machine learning models to predict TVB-N to assess the freshness of lamb was demonstrated. The theoretical basis of the freshness analysis method of other meat product quality is proved.
[Read full article] [Citation]
