Re-thinking the Prediction of Construction Hazard Identification Based on Multimodal Data

Predicting hazard identification performance (HIP) is important in improving engineering project management. However, prediction based on unimodal data may overlook certain important information. Therefore, this study aims to predict HIP using multimodal data. After an experimental study, we found that (1) the prediction results of hazard identification change over time and the optimal time segments for predicting the identification performance of different hazard types are different; (2) the identification of electrical-related and fire-related hazards requires more attentional resources and longer time than fall-related hazards; (3) the physiological mechanism of hazard identification is that the occipital lobe of the brain processes visual information first, while the occipital lobe of the brain transmits visual information later. This study further explores the basic principles of hazard identification and provides a reference for improving safety management in engineering projects.