PDFM Geographic Index Dataset

This dataset is real data released by Google Research in 2024 for evaluating Population Dynamics Based Embeddings. It contains rich summary information of human behavior captured from maps, search trend summaries, and environmental factors such as weather and air quality.General Geospatial Inference with a Population Dynamics Foundation Model".

The dataset contains 3 files:

• conus27 (Interpolation, Super-resolution, and Extrapolation): The conus27 file is a versatile dataset that supports tasks involving interpolation (filling in gaps), super-resolution (predicting at finer spatial scales), and extrapolation (projecting data over large missing areas). The file includes detailed geolocation information (place, county, state, latitude, longitude) and key population health indicators, as well as geographic features such as tree cover, elevation, and nighttime lights.

• Prediction: The model’s ability to predict time is demonstrated using two datasets:
  • county_unemployment.csv: Contains county-level unemployment data from 1990 to 2024, allowing users to track employment trends over time.
  • zcta_poverty.csv: This file provides annual poverty estimates at the ZIP Code Tabulation Area (ZCTA) level from 2011 to 2022, providing insights into economic and social changes at a finer spatial scale.

The Google Research team used graph neural networks to model the complex relationships between these data and locations, and combined the PDFM model with the most advanced forecasting base model TimesFM to predict unemployment and poverty rates, achieving superior performance.