Abstract

More than half of the world's population lives in urban areas and, according to United Nations projections, by 2050 up to 68% of the world's inhabitants will reside in cities. Along with this increase, there are high rates of growth in urban poverty. One of the most evident manifestations of urban poverty in developing countries is the proliferation of slums. Thirty percent of the urban population in developing countries lives in such settlements and this could increase to 60 percent by 2050.

Traditionally, slums and non-slums have been differentiated according to administrative definitions or indicators based on household income. In most countries, official data collection methods rely, through surveys, on headcounts as the basis for deprivation or poverty mapping, and do not provide detailed spatial information on the concentration or location of slum dwellers. However, this is a complex and permanently dynamic phenomenon, so that certain circumstances can make marginalized populations invisible, or worse, exclude them from social policies.

Therefore, in order to identify slums systematically and in a timely manner, it is important to consider variables other than socioeconomic variables, with precise methods for recording and defining them spatially in a coherent manner and in support of geographic targeting. In particular, the use of non-traditional data sources such as satellite images has the potential to identify slums and make visible populations often not identified in traditional surveys.

In this sense, the objective of this session is to present advanced methods based on Artificial Intelligence, Big Data and Machine Learning to map and characterize the spatial extent of slums and poverty.