Urban spatial structure from a street network perspective: mapping street patterns with random forest classification

Street patterns are planar street layouts in a given urban area, which serve as tools for researchers and urban planners to comprehend the structure of urban environments. Nonetheless, the task of mapping street patterns for extensive inter-city studies remains daunting due to the lack of consistency in manual identification methods. With recent technological advancements and data accessibility, new avenues have opened for data-driven techniques in mapping street patterns. This study proposes an innovative framework that employs open data platforms and data processing methods, including network science and supervised machine learning, to map street patterns in cities across the globe effortlessly. Case studies were applied to six cities worldwide and made two key observations from the resulting maps. Firstly, the spatial distribution of street patterns mirrors the urban spatial structure within a city. Secondly, the innate differences between cities become apparent. This study is confident that the novel methodology not only unveils the urban spatial structure across diverse cities but can also be employed to investigate the connection between urban built form and urban activities.