The Legacy of Spatial Apartheid Machine Learning Challenge Data

The dataset was created from data classifying neighbourhoods in South Africa According to 4 neighbourhood types: Wealthy (a combination of the classes Suburb, Smallholdings, Farm), Non-wealthy (combination of Township, Informal area, Collective living Quarters, Village), Non-Residential (combination of Industrial area, Commercial land, Parks and Recreational Areas, Vacant) and Background.

These data are to be used for a machine learning challenge put on by Ro'ya. Please use Application for DAIR Ro'ya data to apply for access.

The specific application the authors created the larger dataset for is to enable researchers and policymakers to quantify the effects of spatial apartheid over time, for the specific purpose of helping to uncover and working to reverse its effects. Those data will be available at a later date.