The aggregate features for inspections and tests are listed and described in Figure 5. At the spatial and temporal scales in Figure 4 there are a total of 63 suchfeatures.For each ve year period ending in 2009 to 2013,the 5-year ACS survey gives us tract-level statistics. Fea-tures include educational achievement (e.g. percent-age of adults who are college graduates), wealth (e.g.percentage of households below the poverty line), andhealth (e.g. percentage of minors that are uninsured).There are 21 such features.The data pipeline is visualized in Figure 6. We used PostgreSQL with the geospatial extension PostGIS for data cleaning and aggregation. Deduplication and dataset assem-bly is done in Python and models are run using the scikit-learn module. The source code is available at the Data Sci-ence for Social Good GitHub repository.