Machine learning-based automated classification of worker-reported safety reports in construction

Limited academic attention has been paid to the applicability of Machine Learning (ML) approaches for analyzing worker-reported near-miss safety reports, as opposed injury at construction sites. Although resource-efficient analysis through ML large volumes such data sites can help guide practitioners in decision-making prevent injuries. The current study addresses this research gap by evaluating relevance quantitative and qualitative methods scaling efficient reporting programs uses an extensive experimentation strategy consisting input processing, n-gram modeling, sensitivity analysis. It first tests proposition that, despite data-quality challenges, high performance different algorithms be achieved automatically classifying textual observations. relies on collected from a real site Kuwait. classification various is evaluated using F1 scores three academically novel but commonly used category labels - "Unsafe Act (UA)," Condition (UC)," "Good Observation (GO)." In addition, practitioner's was utilized assess practical classifiers conventional Logistic Regression (LR) have comparatively score 0.79. However, faced challenges distinguishing between UA UC. Further, reveals that optimal may lose being acceptable human decision-makers. Overall, promising tools data, with low maturity systems find themselves unable leverage scale their systems. A simplified like could identify data-specific future applications.