A food safety targeted sampling decision-making method based on association rule mining and GNNs

Area of Science:

• Food Science
• Data Science
• Risk Management

Background:

• Traditional food safety sampling methods often lack objectivity and precise targeting.
• Subjectivity in sampling can lead to inefficient resource allocation and reduced effectiveness in identifying risks.

Purpose of the Study:

• To develop a targeted sampling decision-making method for food safety.
• To address the limitations of subjectivity and low targeting in current sampling practices.
• To enhance the efficiency and accuracy of food safety monitoring.

Main Methods:

• Constructed a food decision-making factor reasoning module using association analysis.
• Employed an improved frequent pattern growth algorithm to mine food factor association rules.
• Developed a decision-making support module utilizing a graph neural network for sampling frequency.
• Applied the CRITIC-TOPSIS method to determine the sampling sequence of hazardous substances.

Main Results:

• Generated decision-making results for sampling frequency and hazardous substance order.
• Validated the method using nationwide and provincial data for processed grain products (2020-2022).
• Demonstrated the wide applicability and effectiveness of the proposed targeted sampling method.

Conclusions:

• The proposed targeted sampling method significantly improves objectivity and targeting in food safety.
• The integration of association analysis, graph neural networks, and CRITIC-TOPSIS offers a robust framework for sampling decisions.
• This approach provides a valuable tool for enhancing food safety surveillance and risk management.