Research on safety risk assessment model of construction engineering based on attention mechanism and graph neural network

This paper profoundly studies the construction engineering safety risk assessment model based on attention mechanism and graph neural network, aiming at improving the accuracy and timeliness of construction site safety risk early warning. The comprehensive evaluation of multi-dimensional and multi-level risks of construction projects is realized by constructing an evaluation model that combines attention mechanism and graph neural network. In terms of data analysis, this paper uses the historical data of several actual construction projects as training and test samples, covering many key risk areas such as construction period, quality, and capital. The experimental results show that the average accuracy rate of the model on the test set reaches 92.3 %, which is about ten percentage points higher than the traditional risk assessment method, showing excellent performance advantages. Through experiments, we prove that the average accuracy of the model in the test set is 92.3 %, which is about 10 percentage points higher than the traditional risk assessment method. When predicting high-risk areas, the accuracy rate is as high as 95.6 %, which can provide more accurate risk warning information for project managers, show excellent performance advantages, and help to more effectively prevent risks in advance. To sum up, the construction safety risk assessment model based on the attention mechanism and graph neural network proposed in this study not only enriches the theoretical system of construction safety risk assessment but also provides a scientific and efficient risk management tool for practical engineering projects, which has important theoretical significance and application value.