Evaluating a data-driven approach for choice set identification using GPS bicycle route choice data from Amsterdam

Specifying the choice set for travel behaviour analysis is a non-trivial task. Its size and composition are known to influence the results of model estimation and prediction. Most studies specify the choice set using choice set generation algorithms. These methods can introduce two types of errors to the specified choice set: false negative (not generating observed routes) and false positive (including irrelevant routes). Due to increased availability of revealed preference data, like GPS, it is now possible to identify the choice set using a data-driven approach. The data-driven path identification approach (DDPI) combines all unique routes that are observed for one origin-destination pair into a choice set. This paper evaluates this DDPI approach by comparing it to two commonly used choice set generation methods (breadth-first search on link elimination and labelling). The evaluation considers the three main purposes of choice sets: analysis of alternatives in the choice set, model estimation and prediction. The conclusion is that the DDPI approach is a useful addition to the current choice set identification methods. The findings indicate that in analysing alternatives in the choice set, the DDPI approach is most suitable, as it reflects the observed behaviour. For model estimation the DDPI approach provides a useful addition to the current choice set generation methods, as it provides insights into the preferences of individuals without requiring network-data for additional information or generating routes. In terms of prediction, the DDPI approach is not suitable, as it is not able to perform well with out-of-sample data.