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

2025-08-25

AI Summary

Urban time-series forecasting faces a data imbalance challenge: only a few cities possess long-term, dense historical records, while most suffer from sparse data—leading to negative transfer in cross-city knowledge adaptation. To address this, we propose a retrieval-guided selective knowledge transfer framework that combines semantic-aligned source sequence retrieval with large language model (LLM)-based structured reasoning to precisely identify and transfer patterns relevant to the target city, thereby mitigating noise-induced degradation. Our contributions include: (1) a subsequence-aware temporal encoder; (2) a retrieval-augmented LLM inference architecture; and (3) an inference distillation technique tailored for lightweight open-source models. Evaluated on parking datasets from Singapore, Nottingham, and Glasgow, our method significantly outperforms state-of-the-art forecasting and transfer learning baselines, demonstrating strong efficacy, interpretability, and deployment efficiency in low-resource urban settings.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

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.

Problem

• Addresses data imbalance in urban time series forecasting
• Selectively transfers knowledge from data-rich to data-scarce cities
• Prevents negative transfer through retrieval-guided reasoning

Innovation

• Selective knowledge transfer via target-aware retrieval
• Retrieval-guided reasoning with large language models
• Distillation into compact open model via fine-tuning