To address specific problems in civil engineering management, such as inefficient feature extraction from massive multi-source data, delayed risk warning, and rigid resource scheduling, this paper proposes an innovative method integrating machine learning. First, a deep autoencoder is used to perform nonlinear dimensionality reduction and denoising on the original data. Next, a graph neural network is used to extract the spatiotemporal correlation features of engineering entities. Then, a multimodal feature fusion strategy based on an attention mechanism is designed. Finally, a deep reinforcement learning decision model is constructed to achieve adaptive strategy generation. Experiments show that this method achieves a silhouette coefficient of 0.790 in bridge feature extraction, an accuracy of 88.0% in risk warning, and a cumulative reward improvement to 0.894 in dynamic scheduling. The study demonstrates that this fusion framework can effectively improve the intelligence level and decision-making accuracy of engineering management.
Jianbo He (Thu,) studied this question.