Studies using machine learning and satellite imagery reveal that shanty towns may hold more value than previously estimated

All the while, official data has underestimated the number of, and the economic value laden in, slums, suggest two recent studies based on machine learning algorithms and satellite imagery.

The reports—jointly prepared by Sanford School of Public Policy at Duke University, impact investment firm Omidyar Network, North Carolina State University, and the University of North Carolina—were released last month at IIM-Bangalore, an initial partner of the research.

Titled ‘Studying the Real Slums of Bengaluru’ and ‘Characterising Irregular Settlements Using Machine Learning and Satellite Imagery’, the studies focus on Bengaluru, India’s fastest growing city in the last decade, but have implications for the entire country.

Among the findings are that official records vastly underreport the number of slums, the settlements aren’t transitory in nature, living conditions vary across slums and technology must be a part of the solution. The study pegged the average value of a slum at ₹15 lakh and tracked 4,500 slum households for over seven years.