Using Machine Learning and Computer Vision to Increase Property Tax Collection in the DRC

This project involves the construction of a property valuation roll for Kananga, a city of roughly 1 million inhabitants in the Democratic Republic of the Congo (DRC). To calculate the value of the 48,000 properties in Kananga, the researchers propose to use machine learning models trained on survey characteristics of properties and neighbourhoods. The researchers rely on 1,654 property values estimated by government land surveyors during in-person property appraisal visits and combine these data with property characteristics from surveys or extracted from photographs and uses machine learning or computer vision models to predict the values of the remaining 43,508 properties in the city.