Milk Source Identification and Milk Quality Estimation Using an Electronic Nose and Machine Learning Techniques.

Area of Science:

• Agricultural Science
• Sensor Technology
• Analytical Chemistry

Background:

• Milk quality assessment relies on accurate identification of dairy sources and quantification of key components like fat and protein.
• Traditional methods can be time-consuming and costly, necessitating the development of rapid, non-destructive analytical techniques.
• Electronic nose (E-nose) technology offers a promising approach for analyzing volatile organic compounds associated with milk's origin and composition.

Discussion:

• The study developed a cost-effective E-nose with seven metal oxide semiconductor sensors for milk analysis.
• Machine learning algorithms, including Support Vector Machine (SVM) and Random Forest (RF), were employed for classification and regression tasks.
• Feature fusion and dimensionality reduction techniques like Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) were utilized to enhance model performance.

Key Insights:

• The SVM model, utilizing fused E-nose and Dairy Herd Improvement (DHI) data after LDA, achieved 95% accuracy in identifying milk sources (dairy farms).
• Random Forest (RF) models demonstrated superior performance in estimating milk fat (R² = 0.9399) and milk protein (R² = 0.9301) content.
• The developed E-nose system provides a robust, non-destructive method for improving the accuracy of milk quality parameter estimation.

Outlook:

• Further research could explore expanding the sensor array and incorporating advanced machine learning models for more comprehensive milk analysis.
• Integration of E-nose technology into real-time quality control systems on dairy farms could streamline operations.
• Validation of the E-nose system across diverse milk production environments and breeds will be crucial for widespread adoption.