Spatiotemporal Epidemic Control Strategy Framework with Shenzhen Mobility Data

Shang Wang's research dataset, published on figshare in April 2026, presents a framework for epidemic control under incomplete observations. The 94.6 MB repository includes code, models, and data supporting a three-stage method integrating observation modeling, state reconstruction, and multi-agent reinforcement learning. Experiments using community-level mobility data from Shenzhen demonstrate the framework's performance in reducing infection costs.

Use Cases

• Training hybrid ODE-graph-temporal networks for infection state reconstruction based on the described ODE-DynNet model.
• Developing multi-agent reinforcement learning policies for coordinated regional epidemic control based on the described flow-based reward.
• Benchmarking epidemic reconstruction methods under non-steady partial observability conditions described in the paper.
• Simulating intervention outcomes using a metapopulation SEIQR model extended for heterogeneous observation gaps.

Strengths

• Framework demonstrated an 88% reduction in infection costs compared to direct policy optimization under incomplete observations.
• Includes multiple file formats (CSV, NPY, SHP, XLSX) suggesting structured data, code, and geospatial components.
• Experiments are validated using real-world community-level mobility data from a specific city (Shenzhen).

Limitations

• Row count and specific column definitions are unknown, limiting suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The primary focus is on a methodological framework; the raw observational data may not be fully included.

Provenance

Source

Shang Wang via figshare.

Collection Method

Likely generated from simulations and processing of community-level mobility data, as described in the associated research.

Freshness

Last updated 2026-04-11 04:35:58

Geography

Shenzhen, China (mentioned as the source for community-level mobility data).