Predicting occupational injury causal factors using text-based analytics: A systematic review.

Area of Science:

• Occupational Safety and Health
• Data Science
• Computer Science

Background:

• Workplace accidents result in significant human and financial losses.
• Occupational injury reports contain valuable narrative data for analysis.
• Extracting and analyzing this narrative data is crucial for understanding and preventing incidents.

Approach:

• A systematic review of text mining and Natural Language Processing (NLP) applications in occupational injury analysis was conducted.
• Searches across Scopus, PubMed, and Science Direct identified 27 relevant original studies using machine and deep learning models.
• The Preferred Reporting Items for Systematic Review (PRISMA) guidelines were followed.

Key Points:

• Various machine learning (K-means, Naïve Bayes, SVM, Decision Tree, KNN) and deep learning models are used for predicting occupational injuries.
• Deep neural networks are employed for classifying accident types and identifying causal factors.
• There is limited use of deep learning models for extracting injury report narratives, attributed to their recent development.

Conclusions:

• Natural Language Processing (NLP) and text analytics offer significant potential for occupational injury research.
• Future research should focus on improving data balancing techniques and developing automated decision-support systems using deep learning-based NLP models.
• Advancing the application of deep learning in occupational safety and health decision-making is recommended.