SIMULTANEOUS MULTIMODAL DEMAND IMPUTATION AND FORECASTING VIA GRAPH-GUIDED GENERATIVE NETWORK

Accurate perception and prediction of multimodal transport demand are crucial for effective transport management, allowing for services optimization based on historical and future demand. Missing data remains a common challenge to multimodal transport demand analytics, and the potential benefits of knowledge sharing among different modes for simultaneous imputation and forecasting have not been thoroughly investigated, which is tackled by the proposed Graph-guided Generative Imputation and Forecasting Network (GIF) in this work. GIF is constructed based on the Generative Adversarial Network with a Generator to generate missing values and future demand simultaneously and a Discriminator to distinguish synthetic and true data. An Encoder-Decoder framework is employed to reconstruct the generated data to the original data to ensure the important information is preserved. Spatiotemporal features of each mode demand are captured via Transformer-encoder layers while the knowledge sharing among multiple modes is facilitated by graph-guided feature fusion of different modes. The proposed method is evaluated on three real-world transport datasets, demonstrating its potential in addressing the forecasting task with missing data in multimodal transport systems. This study provides insights into the effectiveness of knowledge sharing among modes and joint imputation and prediction in improving the accuracy of multimodal demand imputation and prediction.