Advancing occupational exposure models: insights from a case study.

Area of Science:

• Occupational Health
• Environmental Science
• Data Science

Background:

• Current occupational exposure assessment models are primarily mechanistic, with variable performance.
• Advanced approaches like machine learning (ML) and Bayesian Networks may enhance prediction accuracy.

Purpose of the Study:

• To compare the performance of five ML or hybrid models against an existing mechanistic model for asbestos removal exposure.
• To evaluate the potential of advanced modeling techniques in occupational exposure assessment.

Main Methods:

• Five ML/hybrid models (Multiple Linear Regression, Random Forest, Gradient Boosting Machines, Bayesian Network, Neural Network) were developed.
• These models used the same dataset and determinants as the original mechanistic "asbestos removal exposure assessment tool."

Main Results:

• Random Forest and Gradient Boosting Machines demonstrated the highest accuracy.
• Bayesian Network, Multiple Linear Regression, the original model, and Neural Network followed in performance.
• A trade-off between accuracy and transparency was observed in ML models.

Conclusions:

• ML and hybrid models show potential for developing more accurate occupational exposure assessment tools.
• Bayesian Networks offer a transparent approach by combining expert knowledge and data.
• Further consideration of model limitations is needed for regulatory implementation.