Trip purpose is a crucial element of transportation-related data; however, it is often missing from automatically collected datasets, such as GPS records. Therefore, trip purpose must be inferred from various trip-related attributes to use transportation data effectively. This study developed a trip purpose inference model that incorporates functional zones that vary with time and age. First, these dynamic time–age functional zones were extracted from check-in data to serve as key features. Next, two inference models were constructed, one based on the extreme gradient boosting algorithm and the other on the long short-term memory network, using household and personal travel survey datasets that both included trip purpose information. The developed model was subsequently applied to a daily life movement dataset comprising open real-world data aggregated from mobile phone GPS locations without explicit trip purpose information. The results indicated a purpose-matching rate of 68% for home-to-work trips and 74% for work-to-home trips compared with the movement type information provided in the real-world dataset. These findings suggest that the proposed approach has significant potential for inferring trip purpose from large-scale transportation datasets.
Lee et al. (Mon,) studied this question.