Abstract: Predicting passenger demand is challenging due to nonlinear and dynamic demand patterns. The spatiotemporal dependencies among various zones need to be taken into consideration to improve the prediction accuracy. To address the problem of passenger demand prediction, a GCN-LSTM model is established by combining the graph convolutional network (GCN) and long short-term memory (LSTM). Based on the analysis of the spatiotemporal correlation among various urban areas, passenger demand maps are constructed with the help of the dynamic time warping (DTW) algorithm and the spatial dependencies of the demand map captured using GCN. The LSTM-based encoder is used to capture the temporal dependencies of the zones. The LSTM-based decoder is used to carry out the multi-step predictions of passenger demand for multiple zones simultaneously. The comparison experiments show that the proposed GCN-LSTM model outperforms the traditional models in terms of the root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE), which verifies the prediction accuracy of the proposed model.