Route Choice Identification and Selection from Sparse Floating Car Data Sets

The paper deals with the opportunities and difficulties to exploit large sets of sparse floating car data for modeling purposes, more specifically for route choice analysis. A methodology is introduced for path identification and selection. It explores all possible routes between an origin-destination pair starting from a set of sparse observed vehicle positions; it identifies the most likely routes for each trip and finally selects a limited set of representative paths that appear significantly different on the road network model. Finally, an application is presented on a set of 62 routes between one origin-destination pair, selected from a database of several million of trips tracked in the metropolitan area of Rome. The corresponding set of representative paths is shown which provides the best balance of complexity and accuracy in representing users’ behavior on the road network model.