Decoding urban inequality : the applications of machine learning for mapping inequality in cities of the Global South

Abstract

According to the United Nations, by the year 2050, 68% of the world's population will live in cities. However, the UN also estimates that 1 in 8 people in the world currently live in slums; furthermore, slum populations are growing at a rate of 4.5% per year. Nairobi, the capital of Kenya, is known for having large slum settlements and a high degree of spatial inequality. While slums are expanding at a rapid rate, cities in the Global South lack the crucial data to monitor deepening spatial inequalities. Current urban poverty assessments rely on census data, poverty maps or slum demarcation maps, however, for city planning, these are subject to limitations. It is important to note that while the world is undergoing this immense change in its ecology, we are also experiencing a 'data revolution' which is characterized by a rapid growth in data availability as well as a growing interest in data science techniques such as machine learning (ML). Acknowledging these significant trends, this thesis applies ML to generate useful insights on spatial inequality in Nairobi. The research incorporates data from multiple sources including: census, satellite imagery and data derived from calculations in GIS. The research explored two ML methods. The first method attempted to map living conditions for small areas in the city. Moreover, the second method produced residential typologies or zones for equitable investment and land management in the city. One of the overall aims of the research is to contribute to the wider conversation on how ML may be applied in the realms of policy and city planning in the Global South.