@article{34700, keywords = {Machine learning, Geo-spatial, Neighborhood, Typology, Geotype, Transportation investment, Clustering algorithm}, author = {Natalie Popovich and C Anna Spurlock and Zachary Needell and Ling Jin and Thomas P Wenzel and Colin Sheppard and Mona Asudegi}, title = {A methodology to develop a geospatial transportation typology}, abstract = {

We introduce a methodology to develop a geo-typology (geotype) that categorizes each location in the United States in terms of their main drivers of transportation demand and supply. We develop the first comprehensive set of geotypes for both urban and rural areas across the entire United States. This typology is designed to facilitate national level modeling of multi-modal transportation system's response to alternative investment strategies differentiated across different types of locations. We develop a two-stage clustering procedure to systematically and quantitatively characterize the ways in which locations across the nation are similar or different with respect to their potential response to investment strategies of interest. First, we cluster all 73,057 census tracts, using factor analysis and the CLARA clustering algorithm into “microtypes” based on their street network and economic characteristics. Then we cluster regions (core-basic statistical areas and counties) into “geotypes” using PAM clustering according to their commute configurations, polycentricity and density. The resulting set captures both local and regional variation. These microtypes and geotypes are comparable across all locations, enabling a national level perspective, while maintaining sufficient heterogeneity to support a variety of transportation analyses capturing critical geographic variation.

}, year = {2021}, journal = {Journal of Transport Geography}, volume = {93}, pages = {103061}, month = {05/2021}, issn = {09666923}, doi = {10.1016/j.jtrangeo.2021.103061}, language = {eng}, }