Multiple station passenger flow prediction is crucial but challenging for intelligent transportation systems. Recently, deep learning models have been widely applied in multi-station passenger flow prediction. However, flows at the same station in different periods, or different stations in the same period, always present different characteristics. These indicate that globally extracting spatio-temporal features for multi-station passenger flow prediction may only be powerful enough to achieve the excepted performance for some stations. Therefore, a novel two-step multi-station passenger flow prediction model is proposed. First, an unsupervised clustering method for station classification using pure passenger flow is proposed based on the Transformer encoder and K-Means. Two novel evaluation metrics are introduced to verify the effectiveness of the classification results. Then, based on the classification results, a passenger flow prediction model is proposed for every type of station. Residual network (ResNet) and graph convolution network (GCN) are applied for spatial feature extraction, and attention long short-term memory network (AttLSTM) is used for temporal feature extraction. Integrating results for every type of station creates a prediction model for all stations in the network. Experiments are conducted on two real-world ridership datasets. The proposed model performs better than unclassified results in multi-station passenger flow prediction.
With the increasing aggravations of urban traffic jams, intelligent route planning to reduce traffic time is becoming increasingly critical for drivers. However, traditional route-planning methods, such as graph search and Recurrent Neural Network (RNN)-based methods, struggle to capture the complex dynamics of road networks. Specifically in A*-type methods, routes should be searched instantly on the whole graph for the sake of dynamic changes in edge time consumption. As for RNN-based methods, their shortcomings in capturing long-distance sequence dependencies makes them unsuitable for route planning in metropolises with long routes. Therefore, to better adapt to the complexity of urban traffic, in this paper, an innovative route-planning model called Route Planning with Transformer (RoPT) is proposed. This model is based on the fusion of Graph Convolutional Networks (GCNs) and a Transformer, which uses GCNs for capturing complex spatial dependencies between the current intersection and the destination in a road network. Depending on the self-attention mechanism of the Transformer, the long-distance temporal dependencies between intersections could also be captured effectively. With comprehensive experiments on two real-world traffic datasets, the Porto dataset and the Chengdu dataset, it is demonstrated that RoPT outperforms the best methods, to the best of our knowledge. Moreover, the latent features learned from RoPT are more interpretable.