Occupational Injury Risk Mitigation: Machine Learning Approach and Feature Optimization for Smart Workplace...

Area of Science:

• Occupational Health and Safety
• Data Science
• Machine Learning

Background:

• Accurate forecasting of occupational injury severity is a priority for industries.
• Machine learning offers potential for predictive analysis in occupational safety.

Purpose of the Study:

• To propose a feature-optimized predictive model for anticipating occupational injury severity.
• To identify key features influencing injury severity and compare machine learning model performance.

Main Methods:

• Analysis of 66,405 occupational injury records from the Occupational Safety and Health Administration (OSHA) database.
• Evaluation of five machine learning models: Support Vector Machine, K-Nearest Neighbors, Naïve Bayes, Decision Tree, and Random Forest.
• Application of a feature optimization technique to identify the most impactful features.

Main Results:

• Random Forest model demonstrated superior accuracy and F1-score compared to other models.
• The most significant features identified were 'nature of injury', 'type of event', and 'affected body part'.
• Hyperparameter tuning of the Random Forest model improved prediction accuracy for hospitalization (0.895) and amputation (0.954).

Conclusions:

• Ensemble learning, specifically Random Forest, shows significant potential for accurate occupational injury severity prediction.
• Feature optimization is vital for developing effective predictive models and providing actionable insights for safety practitioners.
• The developed model holds promise for enhancing smart workplace surveillance and informing injury prevention strategies.