Abstract:Spatio-temporal prediction aims to forecast and gain insights into the ever-changing dynamics of urban environments across both time and space. Its purpose is to anticipate future patterns, trends, and events in diverse facets of urban life, including transportation, population movement, and crime rates. Although numerous efforts have been dedicated to developing neural network techniques for accurate predictions on spatio-temporal data, it is important to note that many of these methods heavily depend on having sufficient labeled data to generate precise spatio-temporal representations. Unfortunately, the issue of data scarcity is pervasive in practical urban sensing scenarios. Consequently, it becomes necessary to build a spatio-temporal model with strong generalization capabilities across diverse spatio-temporal learning scenarios. Taking inspiration from the remarkable achievements of large language models (LLMs), our objective is to create a spatio-temporal LLM that can exhibit exceptional generalization capabilities across a wide range of downstream urban tasks. To achieve this objective, we present the UrbanGPT, which seamlessly integrates a spatio-temporal dependency encoder with the instruction-tuning paradigm. This integration enables LLMs to comprehend the complex inter-dependencies across time and space, facilitating more comprehensive and accurate predictions under data scarcity. To validate the effectiveness of our approach, we conduct extensive experiments on various public datasets, covering different spatio-temporal prediction tasks. The results consistently demonstrate that our UrbanGPT, with its carefully designed architecture, consistently outperforms state-of-the-art baselines. These findings highlight the potential of building large language models for spatio-temporal learning, particularly in zero-shot scenarios where labeled data is scarce.

GeoPredict-LLM: Intelligent Tunnel Advanced Geological Prediction by Reprogramming Large Language Models

Real-Time Prediction of TBM Driving Parameters Using Geological and Operation Data<italic/>

UrbanGPT: Spatio-Temporal Large Language Models

GeoLLM: Extracting Geospatial Knowledge from Large Language Models

Spatial-Temporal Large Language Model for Traffic Prediction

Traj-LLM: A New Exploration for Empowering Trajectory Prediction with Pre-trained Large Language Models

Data-driven real-time advanced geological prediction in tunnel construction using a hybrid deep learning approach

TPLLM: A Traffic Prediction Framework Based on Pretrained Large Language Models

Large Language Models Powered Context-aware Motion Prediction in Autonomous Driving

Language Models-enhanced Semantic Topology Representation Learning for Temporal Knowledge Graph Extrapolation

DelayPTC-LLM: Metro Passenger Travel Choice Prediction under Train Delays with Large Language Models

Knowledge Graph Large Language Model (KG-LLM) for Link Prediction

Geolocation Representation from Large Language Models are Generic Enhancers for Spatio-Temporal Learning

Graph Neural Prompting with Large Language Models

Towards Explainable Traffic Flow Prediction with Large Language Models

GATGPT: A Pre-trained Large Language Model with Graph Attention Network for Spatiotemporal Imputation

Large Language Models as Interpolated and Extrapolated Event Predictors

Synergizing LLM Agents and Knowledge Graph for Socioeconomic Prediction in LBSN

StreetviewLLM: Extracting Geographic Information Using a Chain-of-Thought Multimodal Large Language Model

How Can Large Language Models Understand Spatial-Temporal Data?

Exploring the Potential of Large Language Models (LLMs) in Learning on Graphs