Construction safety behavior recognition and classification for port engineering: an empirical study based on machine learning

Port construction sites are characterized by complex and dynamic environments with frequent safety incidents. Traditional manual supervision methods are inefficient, highlighting the urgent need for intelligent behavior recognition technologies. This study proposes a machine learning-based approach for recognizing and classifying safety behaviors in port engineering. By collecting multimodal data from port construction sites, a comprehensive feature system is constructed, incorporating posture features, environmental features, and behavioral sequence features. Support Vector Machines (SVM), Random Forests (RF), and Convolutional Neural Networks (CNN) are used to classify the behaviors. Experimental results demonstrate the effectiveness of the proposed method, with the CNN model achieving an accuracy of 94.321% . This provides strong technical support for intelligent safety management in port engineering construction.