Automation in Construction, 156, 105115
Safety|구충완
Forecasting personal learning performance in virtual reality-based construction safety training using biometric responses
2023International Journals
Research AreaSafety
Professor구충완
AuthorsChoi, D., Seo, S., Park, H., Hong, T., and Koo, C. (Corresponding Author)
Publication Year2023
Journal (Volume, Issue, Pages)Automation in Construction, 156, 105115
AbstractDuring virtual reality-based safety training, it is necessary to immediately and objectively evaluate personal learning performance. In light of this, this study proposed an interpretable machine learning approach for forecasting personal learning performance in VR-based construction safety training using real-time biometric responses. During VR-based safety training ('fall accidents on scaffolding'), eye-tracking and EEG data were collected in real time from 30 participants (i.e., construction workers). The main findings can be summarized as follows. Compared to the full forecast model (FM), the support vector regression algorithm of the simplified forecast model (SM), which considers only principal features as independent variables, demonstrated better prediction performance (i.e., accuracy improvement: 0.087 of mean absolute error, overfitting: one-third level of the FM). This study creates new ground in the field of personalized safety training by enabling real-time monitoring and diagnosis for the cognitive states (i.e., learning performance) of construction workers during VR-based construction safety training.
VR-Based Construction Safety TLearning PerformanceForecast ModelBiometric ResponseMachine Learning AlgorithmConstruction Worker