Machine learning for analysis of occupational accidents registration data

Regardless of the efforts by employers and public organizations to eliminate occupational accidents, the latter are a persistent problem in the construction industry. In the Swedish construction context, there is a desire to identify causes and factors playing a role in work-related accident prevention, as there are large underused databases of collected registrations representing knowledge on causes and context of accidents. The aim of the current contribution is to review the application of machine learning (ML) for the improved prevention of accidents and corresponding injuries, and to identify current limitations. A systematic literature review on the use of ML for analysing accident records data was carried out. In the reviewed literature, ML was applied in the prediction of accidents or their outcome and extracting or identifying causes affecting the risks of injuries. The algorithms used were diverse; Artificial Neural Networks, k-nearest neighbour, logistic regression, Naive Bayesian, Decision Tree and Support Vector Machines, Random Forest, AdaBoost analysis, and Stochastic Gradient Tree Boosting. The results point to the identification of accident-related objects and factors such as placement in time, project characteristics, congested and/or confined workplace, poor visibility, lack of preparation, and safety behaviour. ML combined with data mining techniques such as Natural Language Processing and graph mining, appears to be beneficial in discovering unknown associations between different features and in multiple levels of clusters. However, the research on ML in accident prevention is at an early stage. A consensus regarding the algorithms' performance and prediction accuracy benchmarks, has not been achieved. Future research needs to focus on methods addressing the problem of unbalanced data, improving accident recording process, merging different data sources and research into more attributes (such as risk management), applying deep learning algorithms, and improve the testing accuracy of ML models.