Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Human Mobility during Extreme Events via LLM-Enhanced Cross-City Learning

Yinzhou Tang, Huandong Wang, Xiaochen Fan, Yong Li

Abstract

The vulnerability of cities has increased with urbanization and climate change, making it more important to predict human mobility during extreme events (e.g., extreme weather) for downstream tasks including location-based early disaster warning and pre-allocating rescue resources, etc. However, existing human mobility prediction models are mainly designed for normal scenarios, and fail to adapt to extreme scenarios due to the shift of human mobility patterns under extreme scenarios. To address this issue, we introduce \textbf{X-MLM}, a cross-e\textbf{X}treme-event \textbf{M}obility \textbf{L}anguge \textbf{M}odel framework for extreme scenarios that can be integrated into existing deep mobility prediction methods by leveraging LLMs to model the mobility intention and transferring the common knowledge of how different extreme events affect mobility intentions between cities. This framework utilizes a RAG-Enhanced Intention Predictor to forecast the next intention, refines it with an LLM-based Intention Refiner, and then maps the intention to an exact location using an Intention-Modulated Location Predictor. Extensive experiments illustrate that X-MLM can achieve a 32.8\% improvement in terms of Acc@1 and a 35.0\% improvement in terms of the F1-score of predicting immobility compared to the baselines. The code is available at https://github.com/tsinghua-fib-lab/XMLM.

Predicting Human Mobility during Extreme Events via LLM-Enhanced Cross-City Learning

Abstract

The vulnerability of cities has increased with urbanization and climate change, making it more important to predict human mobility during extreme events (e.g., extreme weather) for downstream tasks including location-based early disaster warning and pre-allocating rescue resources, etc. However, existing human mobility prediction models are mainly designed for normal scenarios, and fail to adapt to extreme scenarios due to the shift of human mobility patterns under extreme scenarios. To address this issue, we introduce \textbf{X-MLM}, a cross-e\textbf{X}treme-event \textbf{M}obility \textbf{L}anguge \textbf{M}odel framework for extreme scenarios that can be integrated into existing deep mobility prediction methods by leveraging LLMs to model the mobility intention and transferring the common knowledge of how different extreme events affect mobility intentions between cities. This framework utilizes a RAG-Enhanced Intention Predictor to forecast the next intention, refines it with an LLM-based Intention Refiner, and then maps the intention to an exact location using an Intention-Modulated Location Predictor. Extensive experiments illustrate that X-MLM can achieve a 32.8\% improvement in terms of Acc@1 and a 35.0\% improvement in terms of the F1-score of predicting immobility compared to the baselines. The code is available at https://github.com/tsinghua-fib-lab/XMLM.

Paper Structure

This paper contains 22 sections, 17 equations, 6 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Problem Definition
  4. Method
  5. Overall Framework
  6. RAG-Enhanced Intention Predictor
  7. Intention Translator and Predictor
  8. Intention-CLIP
  9. Trajectory Retrieval
  10. LLM-based Intention Refiner
  11. Intention-Incorporated Prompt
  12. Event-Level-Awaring Soft Prompt
  13. Intention-Modulated Location Predictior
  14. Experiments
  15. Datasets
  16. ...and 7 more sections

Figures (6)

  • Figure 1: Analysis of mobility prediction performance of existing algorithms trained in the normal scenario (i.e., DeepMove and Flashback) as well as cross-city transfer learning algorithm (i.e., CHAML), and the cause of their underperformance.
  • Figure 2: Overall framework of X-MLM
  • Figure 3: Information flow in our framework, in which $\mathbb{D}^X_S$, $\mathbb{D}^N_S$, $\mathbb{D}^X_T$, $\mathbb{D}^N_T$ refers to the data in extreme scenarios in the source cities, data in normal scenarios in the source cities, data in extreme scenarios in the target cities, and data in normal scenarios in the target cities, respectively.
  • Figure 4: Illustration of Intention-CLIP
  • Figure 5: Top 5 weights of vocabularies in prototypes.
  • ...and 1 more figures

Theorems & Definitions (2)

  • DEFINITION 1: Location
  • DEFINITION 2: Human Mobility Trajectory