Related Experiment Video
Updated: Jul 11, 2025
Evaluation of the Efficacy of Organic Peroxyacids for Eradicating Dairy Biofilms Using an Approach Combining Static and Dynamic Methods
Published on: December 9, 2022
A data-driven approach for prioritising microbial and chemical hazards associated with dairy products using
Gopaiah Talari1, Rajat Nag2, John O'Brien3
1Creme Global, 4th Floor, The Design Tower, Trinity Technology & Enterprise Campus, Grand Canal Quay, Dublin 2 D02 P956, Ireland; University College Dublin, School of Biosystems and Food Engineering, Belfield, Dublin 4, Ireland.
This study uses machine learning to classify dairy product food safety alerts, identifying key chemical and microbial contaminants. Models achieved up to 98% accuracy in classifying serious food safety risks for timely intervention.
Area of Science:
- Food safety analysis
- Data-driven risk assessment
- Machine learning applications in food science
Background:
- Dairy products are susceptible to chemical and microbial contamination, posing significant public health risks.
- Existing food safety alert systems require efficient methods for prioritizing and classifying hazards.
- Databases like RASFF and GEMS provide valuable data for understanding contaminant prevalence and severity.
Purpose of the Study:
- To prioritize microbial and chemical hazards in dairy products based on occurrence and severity.
- To develop and validate machine learning models for classifying the severity of food safety alerts.
- To identify key features influencing the performance of machine learning models in food safety classification.
Main Methods:
- Utilized exploratory data analysis on RASFF and GEMS food contaminant databases.
- Applied machine learning algorithms including decision trees, random forests, k-NN, LDA, and SVM.
- Validated models on an external dataset of RASFF alerts for chemical contaminants in dairy products.
Main Results:
- Identified priority microbial hazards: Listeria monocytogenes, E. coli, Salmonella, etc.
- Prioritized top ten chemical hazards: nitrate, lead, arsenic, ochratoxin A, etc.
- Achieved classification accuracy up to 98% for 'serious' vs. 'non-serious' alerts and 95.1% on external data.
Conclusions:
- Machine learning models demonstrate robustness in classifying dairy food safety alerts for chemical contaminants.
- Key features like Reference dose and substance amount significantly impact model performance.
- Findings support the development of accurate models for timely food safety interventions.
