Best Precision–Recall Confidence Threshold and F-Measure to Determine Quality of Camel Meat by Support Vector Regression Based Electronic Nose

Abstract

AbstractMaintaining fresh quality of camel meat and devising an effective validation instrument were the goals of this project. The minced samples were packed in bags with five different atmospheres and stored 20 days at 4 °C. Head space gas analysis and total viable count of bacteria were performed as references. Meat samples were tested with an electronic nose machine through dynamic sampling. Principal component analysis technique resulted in two distinct clusters in consistence with reference methods. Carbon dioxide was the best modified atmosphere to retain meat freshness. Support vector regression was trained with different confidence thresholds.  The best precision–recall and F-measure values were obtained at threshold 0.5 that are promising to avoid false-positive and false-negative diagnoses which would be very crucial for regulatory decision-making organizations.