A Data-Driven Recommendation System for Construction Safety Risk Assessment

Abstract

Subjectivity and uncertainty of risk assessment (RA) procedures can be improved by replacing guesswork with data-driven approaches such as machine learning (ML). Although a plethora of ML prediction techniques have been introduced to improve the reliability of RA procedures, the utilization of ML-based recommendation systems that can leverage data from multiple aspects has remained unexplored. In this study, a novel RA recommendation system (RARS) was developed to achieve more reliable, objective, and inclusive safety decisions that can prioritize hazard items and formulate related risky scenarios. To this end, a semisupervised graph representation learning framework, node2vec, was utilized to receive semantic and dependency information from safety records to recommend the components of potential accident scenarios (hazards, hazardous cases, dangerous activities, and risky behaviors) based on the given decision objective. The RARS’s ability to provide flexible and user-oriented safety recommendations was explored on a real-life construction accident data set. This allows construction safety practitioners to dynamically evaluate possible risky scenarios with details regarding different influential risk factors and accordingly devise more reliable site safety strategies and relevant policies. The proposed RARS, through its adoption of the graph representation learning-based recommendation model, has the potential to advance hazard identification and risky scenario formulation during the risk analysis and evaluation stages for three reasons: first, a relation-aware representation data set is structured while assigning each hazard item to the project, related safety features, and different construction occupations; second, it allows flexible configuration of the system input based on different decision objectives by the construction professionals; and third, it provides data-driven recommendations by learning the relationship between the characteristics of different safety data collected across various projects while considering the project similarities in terms of the shared safety attributes. The proposed RARS can identify patterns and relationships in construction safety data sets to generate suggestions and recommendations, even in the absence of explicit labels or outcomes. RARS can suggest relevant hazards, hazardous cases, dangerous activities, and risky behavior items, considering the safety features shared among different projects and construction occupations. This facilitates its constant utilization during the procedure of formulating different safety scenarios that are often performed based on experience-driven guess works, while there may be incomplete or missing data, which is a common occurrence in RA procedures.