Early detection of properties at risk of blight using spatiotemporal data

Urban blight is a domino effect phenomenon: properties first fall into disrepair, then land values decline, and finally home abandonment and vacancy follows. This effect spreads from one home to another in the neighborhood, depressing values of nearby properties [8]. In partnership with the City of Cincinnati Office of Performance and Data Analytics and their Department of Buildings & Inspections, we used geographical data from the city and historical data on home inspections to train a Machine Learning model to provide proactive suggestions for property inspections targeted at catching blight early. Our best model reaches a precision of 70% for the top 6,000 predictions. This is a significant improvement over the discovery rate of the current approach, where 60% (in 2015) of citizen complaints result in the discovery of code violations. While our model can have a huge impact in tackling the blight problem, without field validation, the model can potentially have unintended consequences and ethical issues, such risks are being taken into account for the development of the project.