Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach

2025-05-23

AI Summary

In spatiotemporal forecasting for intelligent transportation, weather prediction, and urban planning, existing methods suffer from three key limitations: insufficient multimodal fusion, severe causal confounding, and high computational complexity. To address these, we propose a novel dual-branch causal inference architecture. Our approach introduces the first cross-modal attention–gated mechanism for collaborative modeling of heterogeneous data sources, and designs a GCN-Mamba hybrid encoder that jointly captures spatial dependencies and long-range temporal dynamics while preserving causal interpretability. Evaluated on four real-world benchmarks, our method consistently outperforms nine state-of-the-art models: it achieves up to 9.66% higher prediction accuracy and reduces inference latency by 17.37%–56.11%. The framework thus delivers a principled trade-off among predictive performance, causal transparency, and computational efficiency.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Causal Learning
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Spatio-temporal prediction plays a crucial role in intelligent transportation, weather forecasting, and urban planning. While integrating multi-modal data has shown potential for enhancing prediction accuracy, key challenges persist: (i) inadequate fusion of multi-modal information, (ii) confounding factors that obscure causal relations, and (iii) high computational complexity of prediction models. To address these challenges, we propose E^2-CSTP, an Effective and Efficient Causal multi-modal Spatio-Temporal Prediction framework. E^2-CSTP leverages cross-modal attention and gating mechanisms to effectively integrate multi-modal data. Building on this, we design a dual-branch causal inference approach: the primary branch focuses on spatio-temporal prediction, while the auxiliary branch mitigates bias by modeling additional modalities and applying causal interventions to uncover true causal dependencies. To improve model efficiency, we integrate GCN with the Mamba architecture for accelerated spatio-temporal encoding. Extensive experiments on 4 real-world datasets show that E^2-CSTP significantly outperforms 9 state-of-the-art methods, achieving up to 9.66% improvements in accuracy as well as 17.37%-56.11% reductions in computational overhead.

Problem

• Inadequate fusion of multi-modal information for prediction
• Confounding factors obscuring causal relations in data
• High computational complexity of spatio-temporal models

Innovation

• Cross-modal attention for multi-modal fusion
• Dual-branch causal inference approach
• GCN-Mamba for efficient spatio-temporal encoding