SARN: Structurally-Aware Recurrent Network for Spatio-Temporal Disaggregation
This addresses the challenge of using aggregated open data for downstream AI/ML systems, with incremental improvements in a domain-specific context.
The paper tackles the problem of disaggregating spatio-temporal data from coarse, irregular partitions (e.g., census tracts) to fine-grained ones (e.g., city blocks), proposing the Structurally-Aware Recurrent Network (SARN) which integrates spatial attention with GRUs. The result shows SARN outperforms neural models by 5% and 1% and heuristic methods by 40% and 14% on two mobility datasets.
Open data is frequently released spatially aggregated, usually to comply with privacy policies. But coarse, heterogeneous aggregations complicate learning and integration for downstream AI/ML systems. In this work, we consider models to disaggregate spatio-temporal data from a low-resolution, irregular partition (e.g., census tract) to a high-resolution, irregular partition (e.g., city block). We propose an overarching model named the Structurally-Aware Recurrent Network (SARN), which integrates structurally-aware spatial attention (SASA) layers into the Gated Recurrent Unit (GRU) model. The spatial attention layers capture spatial interactions among regions, while the gated recurrent module captures the temporal dependencies. Each SASA layer calculates both global and structural attention -- global attention facilitates comprehensive interactions between different geographic levels, while structural attention leverages the containment relationship between different geographic levels (e.g., a city block being wholly contained within a census tract) to ensure coherent and consistent results. For scenarios with limited historical training data, we explore transfer learning and show that a model pre-trained on one city variable can be fine-tuned for another city variable using only a few hundred samples. Evaluating these techniques on two mobility datasets, we find that on both datasets, SARN significantly outperforms other neural models (5% and 1%) and typical heuristic methods (40% and 14%), enabling us to generate realistic, high-quality fine-grained data for downstream applications.