Metal Oxide-Based Sensor Arrays With MultiBoosting Algorithm for Detection of Milk Spoilage Gases and Freshness Assessment

The quality and freshness assessment of milk is crucial for the public health and the detection method using electronic nose (EN) technology is gaining popularity. However, the rapid development of EN technology has been challenged by the sensing performance of commercial gas sensors, the neglect of transient characteristics by conventional response characterization, and the low efficiency of the algorithms. Herein, four excellent metal oxide gas sensors were used to detect simulate milk spoilage volatile gases of ethanol, acetone, and hexanal gases and to obtain multidimensional response data. Grid search (GS) was utilized to automatically optimize deep features extracted by the Daubechies4 (db4) wavelet transform, and then an ensemble algorithm of MultiBoosting was applied to perform qualitative and quantitative analysis of gases. The gas identification accuracy reached 96.77%, and the root mean square error (RMSE) and coefficients of determination ( ${R}^{{2}}$ ) for ethanol, acetone, and hexanal were (2.264 ppm, 0.995), (1.180 ppm, 0.999), and (4.792 ppm, 0.980), respectively. Furthermore, the proposed model was applied to the assessment of milk freshness, demonstrating satisfactory predictive performance. The results indicated that this method holds substantial practical application potential in the detection of milk spoilage gases.