Traffic congestion of large cities through a large-scale online platform

Abstract

Online real-time traffic data services have effectively delivered congestion information to people all over the world and provided great benefits to society. We study urban traffic data of 17 cities from a major online real-time traffic information provider. We sampled the online platform every 5 minutes over twelve months, in total more than 2 million samples covering over 170000 road segments. We use three variables to characterize traffic information for different cities: data availability, duration, and reliability of travel delays. Data availability measures the percentage of real-time information. Travel delays measure the average excess travel time for a given road segments (in travel time per kilometer). Reliability of travel delays is captured by the share of recurring congestion: the larger proportion of non-recurring congestion in total delays, the lower reliability of travel time estimates. We measure traffic data quality by developing methods of filling in missing data and compare that against data provider’s ”historical values;” and by validating against measurements of traffic sensors. Our goal is to make objective assessments of traffic data quality information more transparent and accessible. Future work is critically needed for collecting more case studies of ground truthing especially for developing countries.