Fresh Beef Shelf-Time Estimation Based on Bioimpedance Spectroscopy Measurements and AutoGluon

Meat is an important source of beneficial nutrients for human beings, but expired or rotten shelf meat will become the source of food-borne illness, which means that meat shelf-time estimation (STE) plays an important role in protecting public health. The traditional meat freshness detection methods, such as sensory evaluation, total volatile basic nitrogen (TVB-N) test, and aerobic plate count (APC) test, are important meat STE methods but all have some shortcomings. This paper proposes a novel meat STE method based on bioimpedance spectroscopy (BIS) measurements and AutoGluon for fresh shelf beef. First, a bar four-electrode fixture was designed, and the BIS data were acquired using a commercial impedance analyzer on 42 beef samples for 7 consecutive days at a constant temperature of 8 °C. Then, a beef STE model with seven categories was built using the AutoGluon framework, including a feature-selection preprocessing procedure for the raw BIS data based on genetic algorithm (GA) combining with K-nearest neighbor (KNN). The experimental results showed that the beef STE model achieved an accuracy of 93.69%, outperforming other intelligent algorithms. This paper provides an inexpensive and effective STE approach for shelf beef quality evaluation and will have significant reference value and application prospects for quality evaluation of other fresh meat and fishes.