DWGCN: a route recommendation model based on deepwalk-graph convolutional network

Route Recommendation has been the foundation of trip planning, and although there are many studies in the literature, there are still many challenges. The first one is the problem of how to accurately obtain the spatial features of the road network; the second one is the accessibility of the planned route. To solve these problems, we propose a new path planning model, DeepWalk Graph Convolutional network (DWGCN). The model learns the spatial features of the road network from historical trajectory data by DeepWalk combination with GCN. A Multi-Layer Perceptron component (MLP) is used to estimate the conditional transfer distribution. We finally validate the model with five real datasets, and the experimental results show that our model has some improvements in precision, recall rate and reachability compared with the benchmark model.