Interaction strength prediction based on neural networks of graph scrolls considering geographical features

Accurate interaction strength prediction plays an important role in urban planning and development. Based on the existing national interaction intensity data between cities, the research uses GCN model, ChebNet model and GAT model based on graph convolutional neural network to realize the prediction of urban interaction intensity, and determines the optimal model through accuracy evaluation. Further we introduce geographical features, optimize the optimal GAT model, and finally the graph neural network model considering geographical characteristics is constructed to realize real-time, efficient, and long-term prediction of urban interaction intensity. It is found that there is a strong interaction betweend cities.