Nafe Assafi, M.; Wang, J.; Ma, J. and Cotton, J. (2026) Assessing the impact of eye-tracking features on construction hazard identification skills using virtual reality and machine learning. Construction Innovation, 26(5), pp. 1614-1635. ISSN 1471-4175
Abstract
Purpose – Identifying hazards in construction sites is essential for ensuring safety. This study aims to propose a machine learning-based approach to identify the most effective set of eye-tracking features for assessing hazard identification skills on construction sites. By determining which features best contribute to hazard identification across various hazard types, the approach seeks to improve workplace safety. Design/methodology/approach – Four hazard types and 11 eye-tracking features were identified, and 18 strategies were developed to evaluate hazard identification skills. Support vector machine (SVM) and artificial neural network (ANN) models were applied to assess these strategies. Virtual reality simulations were used to gather eye-tracking data and evaluate hazard identification. Findings – The SVM and ANN models effectively identified the most impactful set of eye-tracking features for each hazard type individually and all hazard types together. The results indicated that the proposed approach can accurately assess hazard identification skills in construction, offering a data-driven means of reducing injuries. Research limitations/implications – This study advances safety research by identifying the most relevant eye-tracking features linked to different types of construction hazards using a machine learning approach. The results can help develop personalized VR-based training and evaluation systems that improve workers' visual attention for hazard identification, leading to safer construction practices. Originality/value – While previous studies have explored eye-tracking features for assessing hazard identification skills, the effectiveness of certain features has been unclear. This study clarifies which set of eye-tracking features is most effective for assessing hazard identification skills based on specific hazard types, contributing to enhanced workplace safety.
| Item Type: | Article |
|---|---|
| Uncontrolled Keywords: | artificial neural network; construction safety; eye-tracking features; hazard identification skills; machine learning; support vector machine; virtual reality |
| Index terms: | effectiveness, virtual reality, hazard identification, safety research, construction safety, machine learning, methodology, artificial neural network, construction site, strategy, injury, workplace safety |
| Subjects: | management, performance management, environmental health, occupational health and safety management, health conditions and diseases, virtual reality, artificial intelligence, work location, modelling and simulation, financial risk, research methods, occupational health |
| Topics: | Health and Safety, Sustainability, Quality Management, Cost Management, Business Strategy, Research Practice, Digital Applications, Site Management |
| Descriptive scope: | 3 PCT |
N.B. Descriptive scope is a count of how many of the five facets of empirical research are indicated by the words used in title, abstract and keywords. It is not intended as a judgement on the research; merely a count of the kind of word we would expect to indicate Phenomenon, Concepts, Theoretical framing, Empirical techniques, Analytical techniques. If all five are present, then a code of “5 PCTEA” will indicate this. If you feel the coding for this record is questionable, we welcome discussion around the terms we matched or the way we categorized them. The facet you would expect may not be coded, or a facet may be coded inappropriately. This can also bear on a larger question, of which facets should be treated as defining in construction management research. Please get in touch, and we will look at it. More details here