Data-driven street segments categorization based on topological properties in urban street networks

The function-based classification (FCS) classifies streets according to their respective requirements in the pre-defined hierarchy of the urban street network (USN). However, a mismatch between the planned and actual performance can often be observed because extensive data-collection or prior local knowledge of the real performance are not always available or are often cost- and resource-consuming. This study proposes a machine learning approach for network-based categorization of street segments (NSC). Measurements derived from network science are computed for each street segment and then clustered based on their topological importance. NSC is then compared with the FCS in order to explore the fine variations in spatial-structural properties of the segments within the existing FCS scheme and to offer opportunities for better planning.