A Foundation Model for Unified Urban Spatio-Temporal Flow Prediction

2024-11-20

arXiv.org

AI Summary

Urban spatiotemporal flow forecasting (e.g., traffic or human mobility) is critical for smart city management, yet existing methods suffer from fragmented modeling of grid-structured versus graph-structured urban data. This paper introduces UniFlow—the first unified foundation model enabling joint representation learning and forecasting across both grid and graph paradigms. Its core innovations include a multi-view spatiotemporal patching mechanism, a dedicated spatiotemporal Transformer architecture, and a spatiotemporal memory retrieval augmentation (ST-MRA) module that facilitates cross-paradigm knowledge transfer and few-shot generalization. UniFlow achieves state-of-the-art performance on both grid-based and graph-based benchmark tasks. Under data-scarce conditions, it reduces mean absolute error (MAE) by up to 23.6% over prior methods. The code and datasets are publicly released.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Urban spatio-temporal flow prediction, encompassing traffic flows and crowd flows, is crucial for optimizing city infrastructure and managing traffic and emergency responses. Traditional approaches have relied on separate models tailored to either grid-based data, representing cities as uniform cells, or graph-based data, modeling cities as networks of nodes and edges. In this paper, we build UniFlow, a foundational model for general urban flow prediction that unifies both grid-based and graphbased data. We first design a multi-view spatio-temporal patching mechanism to standardize different data into a consistent sequential format and then introduce a spatio-temporal transformer architecture to capture complex correlations and dynamics. To leverage shared spatio-temporal patterns across different data types and facilitate effective cross-learning, we propose SpatioTemporal Memory Retrieval Augmentation (ST-MRA). By creating structured memory modules to store shared spatio-temporal patterns, ST-MRA enhances predictions through adaptive memory retrieval. Extensive experiments demonstrate that UniFlow outperforms existing models in both grid-based and graph-based flow prediction, excelling particularly in scenarios with limited data availability, showcasing its superior performance and broad applicability. The datasets and code implementation have been released on https://github.com/YuanYuan98/UniFlow.

Problem

• Unifies grid-based and graph-based urban flow prediction models.
• Introduces a spatio-temporal transformer for complex correlation capture.
• Enhances predictions with adaptive memory retrieval in data-limited scenarios.

Innovation

• Unified grid and graph data processing
• Multi-view spatio-temporal patching mechanism
• Spatio-temporal transformer with memory retrieval