A New Urban Typology Model Adapting Data Mining Analytics to Examine Dominant Trajectories of Neighborhood Change: A Case of Metro Detroit

Abstract

This article develops an integrated methodology to investigate dominant trajectories of neighborhood change that are often confronted in urban studies. Currently, researchers are using k-means cluster analysis to establish diverse neighborhood typologies and principal component analysis (PCA) to identify socioeconomic interactions explaining the neighborhood typologies. Little attention has been given to longitudinal trajectories and dynamics of neighborhood evolution over a long period. Our new model adapts a newly developed dynamic sequential analysis (the weighted minimum edit distance algorithm) in big data analytics to sort and identify dominant trajectories of neighborhood change. Our model also innovatively synthesizes three statistical procedures—k-means, PCA, and analysis of variance—to derive the weight matrix, which naturally integrates the core characteristics of urban neighborhood changes into the sequential reordering. Using the census data in Metro Detroit over five census years (1970, 1980, 1990, 2000, and 2010), this model was tested to identify a unique city's demographic and socioeconomic transition pattern in the past forty years. This model successfully provided a thorough analysis of the neighborhood typologies and exhibited a much-enhanced performance in identifying long-term trajectories of urban evolution.