DescriptionPublic transport smart card data hold vast amount of information on passenger behaviour. However, no information on trip purpose is recorded, hence limiting its use in practice. While several studies have developed methods for estimating trip purpose, estimation accuracy is still a challenge. This study proposes a two-fold methodology for trip purpose inference, which incorporates i) cluster analysis of trip purposes, and ii) trip purpose estimation. The grouping of similar trip purposes through cluster analysis reduces the complexity of the subsequent trip purpose estimation, hence ensuring a better performance. Several methods are applied for the trip purpose estimation, including discrete choice models with the utility function being specified using Bayesian relevance determination. Preliminary results solely based on temporal characteristics are promising. Future work will include also relevant land use characteristics as well as estimation based on other relevant methods, including Random Forests.
|Period||1 Jun 2022|
|Event title||10th Symposium of European Association for Research in Transportation|
|Degree of Recognition||International|