Authors: Jonathan Reades*, King's College London, Jordan de Souza, King's College London, Elizabeth Sklar, King's College London
Topics: Quantitative Methods, Urban Geography, Geographic Information Science and Systems
Keywords: machine learning, neighbourhood change, gentrification, London, random forests
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Bayside A, Sheraton, 4th Floor
Presentation File: No File Uploaded
Recent developments in the field of machine learning offer robust new ways of modelling complex socio-spatial processes at very fine scales and using dozens of variables exhibiting signs of collinearity. Drawing on earlier empirical and theoretical attempts to understand and model neighbourhood change, we examine how the ‘random forests’ approach can help us to address challenges in both classifying and predicting neighbourhood change over time.
We will present an analysis of socio-economic transitions in London neighbourhoods between 2001 and 2011 using a range of open data from the Census and other sources. These are used a training data set to help us predict those neighbourhoods most likely to demonstrate ‘uplift’ or improvement by 2021. We will explore how the results shed light on different types of change and briefly discuss how machine learning approaches differ from traditional quantitative geography methods in terms of their aims and objectives. The implications of such modelling for our understanding of gentrification will be discussed.