The purpose of this research lies in two aspects. The first is to carry out an in-depth comparison of the clustering results obtained from different methods, so as to find the “optimal” clustering method. The second is to conduct cluster analysis of Chicago utilizing the census tracts data from three years 1950, 1980 and 2010, aiming to understand the spatio-temporal transitions of Chicago neighborhoods across the years. This study is both a theoretical investigation and a methodological investigation. The results indicates K-means clustering using principal components proved to be the optimal method in terms of the measures we chose, and the longitudinal change in spatial patterns of Chicago resembled the description of Chicago School theorists, while some brand-new trajectories were also detected.