Integrating Big Data in Metropolitan Regions to Understand Driving Volatility and Implications for Intelligent Transportation Systems

Abstract

Higher driving volatility, e.g., hard accelerations or hard braking, can imply unsafe outcomes, more energy use, and higher emissions. This presentation will demonstrate how large-scale data, increasingly available from sensors, can be transformed into useful knowledge. This is done by creating a framework for combining data from multiple sources and comparing counties/regions in terms of driving volatility of resident drivers. The unique database was created from four sources that include large-scale travel surveys, historical traffic counts from California and Georgia Department of Transportation, socio-demographic information from Census, and geographic information from Google Earth. The database provides a rich resource to test hypothesis and model driving decisions at the micro-level, i.e., second-by-second. The database has 117,022 trips made by 4,560 drivers residing in 78 counties of 4 major US metropolitan areas across two states. They represent significant variations in land use types and populations; all trips were recorded by in-vehicle GPS devices giving 90,759,197 second-by-second speed records. The data integration helps explore links between driving behaviors and various factors structured in hierarchies, i.e., the data are structured at the levels of trips, drivers, counties, and regions. Appropriate hierarchical models are estimated to study correlates of driving performance and to compare traffic performance across regions. The implications of our analysis for intelligent transportation systems will be discussed.

Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.