Automatic prediction of tourist travel trajectory using joint graph convolutional neural network

In order to overcome the problems of low overlap between predicted and actual paths and low TOP-K hit rate in traditional methods, a new automatic prediction method of tourist travel trajectory using joint graph convolutional neural network is proposed. First, model the spatial behaviour of tourists in scenic spots as a road network interaction process. Second, establish a behaviour benchmark model using a homogeneous Markov chain, and construct a multidimensional utility function covering perception, road network, facilities, and passenger flow to support interpretable decision modelling. Then, design a dynamic graph neural network based on modelling results, which encodes trajectories through three channels: Markov priors, static road networks, and dynamic spatiotemporal signals. Finally, achieve trajectory prediction through adaptive gating fusion. The experimental results show that the overlap rate between the predicted path and the actual path of the design method is high, and the TOP-K hit rate reaches 0.823.