Selecting the most relevant variables towards clustering bus priority corridors

This paper proposes a novel framework to identify the most informative variables for clustering bus priority corridors according to their similarities regarding system and operational aspects. Although each bus corridor has its peculiarities, understanding the similarities (e.g., system, physical and operational aspects) between corridors of different regions of the world can help researchers and transit specialists to draw up strategies tailored to improving the traffic in the cities. For that matter, we integrate a novel metric for measuring clustering quality to the omit-one-variable-out-at-a-time selection procedure. The proposed method relies on three steps: (i) collect and preprocess data describing bus corridors; (ii) define the number of clusters to be generated based on a hierarchical approach; and (iii) iteratively group bus corridors, and eliminate less relevant clustering variables. When applied to a dataset comprised of 296 bus priority corridors from 45 countries and described by 44 variables, the proposed framework retained only four variables (i.e., brand and/or logo, station spacing, enhanced stations, and operating speed) responsible for the best stratification of corridors. Four clusters were formed and qualitatively assessed regarding their similarities in terms of system, physical and operational aspects. Corridors were grouped into basic corridors (cluster 1), improved corridors (cluster 2), Bus Rapid Transit (BRT) and Bus with High Level of Service (BHLS) systems (cluster 3), and express, limited-stop services (cluster 4).