Dairy Safety Prediction Based on Machine Learning Combined with Chemicals

Area of Science:

• Food Science
• Machine Learning
• Data Science

Background:

• Public concern over dairy product safety necessitates improved methods for ensuring food security.
• Unsafe dairy products pose significant risks to public health and human lives.
• Existing dairy safety measures and warnings require enhancement through advanced predictive tools.

Purpose of the Study:

• To develop a machine learning-based predictive model for assessing dairy product qualification.
• To identify critical features for accurate dairy safety classification.
• To enhance the reliability of dairy product safety assessments.

Main Methods:

• Utilized 34 common dairy sampling inspection items as initial features.
• Applied feature selection techniques to identify an optimal subset of predictive features.
• Constructed and evaluated various machine learning classification models.

Main Results:

• A predictive model incorporating "total plate count," "water," and "nitrate" demonstrated superior performance.
• The optimized model achieved a sensitivity (SN) of 62.50%, specificity (SP) of 91.67%, and accuracy (ACC) of 72.22%.
• Integrated machine learning algorithms yielded higher prediction accuracy compared to non-integrated approaches.

Conclusions:

• Presents a novel machine learning approach for dairy safety assessment.
• Aims to improve dairy product quality and ensure consumer safety.
• Contributes to mitigating the risks associated with dairy insecurity through predictive analytics.