Predicting trip purpose in a smartphone-based travel survey

Travel Surveys are considered promising candidates to go ‘smart’. Respondents need to be both motivated and competent to correctly report all details of their travels for a specified time period. Location tracking offers options to remove burden and to improve quality of measurement. Adding contextual information, the collected location data may also be input to predictions of travel mode and travel purpose. However, location tracking may be perceived as privacy-sensitive by respondents.

In 2022, Statistics Netherlands again fielded a travel-app assisted experiment in which respondents were asked to use an app with location tracking for thje duration of a week. They were asked to check and, if needed, adjust the stops and trips. The purpose of a trip always had to be provided by the repondent.

In this paper, we evaluate and discuss various machine learning approaches towards predicting the purposed reported by the respondents. Predctions attain a reasonable accuracy but performance depends strongly on the type of trips. The number of visits to a location is one of the strongest predictors.

Zahroh, S., P. Lugtig, Y. Gootzen, J. Klingwort and B. Schouten (2025). Predicting trip purpose in a smartphone‐based travel survey. Discussion paper, Statistics Netherlands, The Hague/Heerlen.