STRATA-TS: Selective Knowledge Transfer for Urban Time Series Forecasting with Retrieval-Guided Reasoning

TL;DR

STRATA-TS is a novel framework that enhances urban time series forecasting by selectively retrieving and reasoning over relevant source data, addressing data imbalance issues with interpretable transfer.

Contribution

It introduces a retrieval-guided reasoning approach with domain-adapted source selection and distillation into a compact model for improved forecasting in data-scarce urban settings.

Findings

Outperforms existing forecasting and transfer baselines

Provides interpretable knowledge transfer pathways

Demonstrates effectiveness across multiple city datasets

Abstract

Urban forecasting models often face a severe data imbalance problem: only a few cities have dense, long-span records, while many others expose short or incomplete histories. Direct transfer from data-rich to data-scarce cities is unreliable because only a limited subset of source patterns truly benefits the target domain, whereas indiscriminate transfer risks introducing noise and negative transfer. We present STRATA-TS (Selective TRAnsfer via TArget-aware retrieval for Time Series), a framework that combines domain-adapted retrieval with reasoning-capable large models to improve forecasting in scarce data regimes. STRATA-TS employs a patch-based temporal encoder to identify source subsequences that are semantically and dynamically aligned with the target query. These retrieved exemplars are then injected into a retrieval-guided reasoning stage, where an LLM performs structured inference over target inputs and retrieved support. To enable efficient deployment, we distill the reasoning process into a compact open model via supervised fine-tuning. Extensive experiments on three parking availability datasets across Singapore, Nottingham, and Glasgow demonstrate that STRATA-TS consistently outperforms strong forecasting and transfer baselines, while providing interpretable knowledge transfer pathways.