Forecasting for short-term ridership is the foundation of metro operation and management. A prediction model is necessary to seize the weekly periodicity and nonlinearity characteristics of short-term ridership in real-time. First, this research captures the inherent periodicity of ridership via seasonal autoregressive integrated moving average model (SARIMA) and proposes a support vector machine overall online model (SVMOOL) which insets the weekly periodic characteristics and trains the updated data day by day. Then, this research captures the nonlinear characteristics of the ridership via successive ridership value inputs and proposes a support vector machine partial online model (SVMPOL) which insets the nonlinear characteristics and trains the updated data of the predicted day by time interval (such as 5-min). Afterwards, to avoid the drawbacks and to take advantages of the strengths of the two individual online models, this research takes the average predicted values of two models as the final predicted values, which are called support vector machine combined online model (SVMCOL). Finally, this research uses the 5-min ridership at Zhujianglu and Sanshanjie Stations of Nanjing Metro to compare the SVMCOL model with three well-known prediction models including SARIMA, back-propagation neural network (BPNN), and SVM models. The resultant performance comparisons suggest that SARIMA is superior for the stable weekday ridership to other models. Yet the SVMCOL model is the best performer for the unstable weekend ridership and holiday ridership. It shows that for metro operation manager that gear toward timely response to real-world unstable and abnormal situations, the SVMCOL may be a better tool than the three well-known models.

Due to different functions and land use types, there are significant differences in ridership patterns among different urban rail transit stations. Considering the characteristics of different ridership and coping with the uncertainty, periodical and stochastic natures of short-term passenger flow, and this paper proposes a novel hybrid methodology that combines the autoregressive integrated moving average (ARIMA) model and wavelet decomposition, which has strong strengths in signal processing, to short-term ridership forecasting. The seasonal ARIMA is used to represent the relatively stable and regular ridership patterns while the wavelet decomposition is used to capture the stochastic or sometimes drastic changing characteristics of ridership patterns. The inclusion of wavelet decomposition and reconstruction provides the hybrid model with a unique strength in capturing sudden change in ridership patterns associated with certain rail stations. The case study is carried out by analyzing real ridership data of Metro Line 1 in Nanjing, China. The experimental results indicate that the hybrid method is superior to the individual ARIMA model for all ridership patterns, but particularly advantageous in predicting ridership at stations often associated with sudden pattern changes due to special events.

Due to data sparsity, insufficient spatial relationships, and the complex spatial and temporal characteristics of passenger flow, it is very challenging to achieve a high prediction accuracy on Origin–Destination (OD) in a large urban rail transit network. This paper proposes a two-stage prediction network GCN-GRU, using a Graph Convolutional Network (GCN) with a Gated Recursive Unit (GRU). The GCN can obtain the adjacency relationship between different stations by selecting the adjacent neighborhoods and interacting neighborhoods of a station and capturing the spatial characteristics of the OD passenger flow. Then, an advanced weighted aggregator is employed to gather important information from the two above-mentioned types of neighborhoods to capture the spatial relationship of the network OD passenger flow and to perceive the sparsity and range of the OD data. On the other hand, the GRU can extract the temporal relationship, such as periodicity and other time-varying trends. The effectiveness of GCN-GRU is tested with a real-world urban rail transit dataset. The experimental results show that whether it is the OD passenger flow matrix of each period (one hour) on weekdays and weekends or the single-pair OD passenger flow between stations, the proposed GCN-GRU models perform better than the benchmark models. This study provides an important theoretical basis and practical applications for operators, thus promoting the sustainable development of urban rail transit systems.