Joint Short-Term Origin-Destination Demand Prediction for Multimodal Transport Systems

Area of Science:

• Transportation Science
• Data Science
• Applied Mathematics

Background:

• Short-term origin-destination (OD) demand prediction is vital for multimodal transportation systems.
• Existing methods struggle with real-time data availability, demand sparsity, high dimensionality, and intermodal influences.

Purpose of the Study:

• To develop a novel model for accurate short-term multimodal transportation OD demand prediction.
• To address challenges of data availability, sparsity, high dimensionality, and intermodal correlations.

Main Methods:

• Proposed a multitask learning and Partial-Differential-based model (PD-MTSOD).
• Incorporated an OD demand learner for real-time demand estimation.
• Utilized hypergraph attention for spatiotemporal feature aggregation.
• Decomposed OD demand and employed partial differential methods to model intermodal correlations.

Main Results:

• PD-MTSOD demonstrated superior performance compared to baseline models in tests on Beijing and New York City multimodal systems.
• Validated the benefits of jointly considering multiple transportation modes.
• Revealed significant correlations between different transportation modes.

Conclusions:

• The PD-MTSOD model offers a reliable approach for short-term multimodal transportation OD demand prediction.
• Jointly analyzing multiple transportation modes enhances prediction accuracy.
• Understanding intermodal correlations is crucial for effective transportation management.