Spatiotemporal Forecasting with LLMs and Knowledge Graph Reasoning

Ongoing project on combining foundation models with graph reasoning for spatiotemporal prediction.

This project studies how large language models and knowledge graph reasoning can improve spatiotemporal forecasting pipelines.

Scope

  • Review methods for spatiotemporal forecasting, LLM integration, and knowledge graph reasoning.
  • Design workflows for using external context in forecasting.
  • Build data curation pipelines from heterogeneous sources.

Current Work

  • Building reproducible preprocessing pipelines.
  • Defining benchmark tasks for context-aware forecasting.
  • Evaluating how knowledge graph signals help forecast quality and interpretability.

Status

In progress (July 2025 - Present).