An Association-Rules Learning Approach to Unsupervised Classification of Street Networks

Street networks (SN) are intricate connections of roads and intersections; there is an enormous amount of available, open source street network (SN) data, such as the worldwide SN hosted by OpenStreetMaps, a popular source of SNs used in SUMO (Simulation of Urban Mobility) models, including the TAPASCologne project (TAPASCP). These SNs are intricate and high in feature dimension, notable are the edges containing data about road name, number of lanes, speed limit, etc. As a result, SN models that utilize this data are robust and highly-detailed. In this paper, we identify five significant ed ge features (SEF) from exploratory data analysis of the TAPASCP, present an unsupervised methodology of classifying bags of edges though an association rules learning approach to SEF aggregation, and measure the success of our methodology with SUMO simulations.