Experimental analyses and clustering of travel choice behaviours by floating car big data in a large urban area

This study introduces a general methodology to process sparse floating car data, reconstruct the routes followed by the drivers, and cluster them to achieve suitable choice sets of significantly different routes for calibrating behavioural models. This methodology is applied to a large set of floating car data collected in Rome in 2010. Results underlined that routes assigned to different clusters are actually very different to each other. Nevertheless, as expected according to Wardrop's principle, the clusters belonging to the same origin-destination have rather similar average route travel times, even if there is a large range between their minimum and maximum values. A focus on drivers' behaviour highlighted their propensity to follow the same route to their usual destination, though the 12% of the drivers switched to an alternative route. However, the analysis conducted over the 1 month of observations did not reveal the existence of any systematic correlation between neither the change of route nor the change of departure time and the travel time experienced the day before.