Future forecast models for regional travel – Combining different data sources in the best way (FRID)
This project investigates how alternative data sources such as mobile phone network data, public transport automatic fare collection data, and GPS trajectories of car traffic, can complement survey data with low response rates, to estimate mode and destination choice models for regional passenger travel using data fusion.
To develop and improve future transport forecast models, several different data sources need to be utilized. Mobile phone network data provides large amounts of data with potentially high representativeness since almost all travellers today carry a mobile phone. Automatic fare collection data, combined with automatic passenger counts, also provides large amounts of data with potentially high representativeness but is limited to public transport. GPS trajectories typically provide very detailed data, such as route choices, but the number of observations is often smaller than that from mobile phone network and fare collection data. Common to mobile phone network data, fare collection data, and GPS trajectories is that these data sources lack information about trip purpose, travel party size, and socioeconomics such as age, gender, car ownership, and driver’s license ownership.
Travel survey data via questionnaires have declining response rates in Sweden (in 2021, the non-response rate was 72% according to Trafikanalys), and there is a high risk of bias, partly because many of those who belong to large households, travel frequently, or have time constraints choose not to respond to the survey. On the other hand, questionnaire surveys provide detailed data on trip purpose, travel party size, and socioeconomics.
Recent research has shown that for long-distance trips (longer than 100 km), survey data can be used in combination with mobile phone network data to estimate a mode choice model, improving the estimation compared to using only mobile phone network data. In models for long-distance trips, the start and end zones are relatively large, which has been advantageous due to the resolution of mobile phone network data. Mobile phone network data can also be used to estimate models for trips shorter than 100 km but may need to be supplemented with other data sources. Regional trips constitute the largest part of passenger travel, and it is important to ensure that we can make accurate forecasts for these in the future as well. Therefore, this project aims to combine different data sources on regional trips in the estimation of a mode and destination choice model. A pilot estimation will be conducted for regional trips in the Norrköping area.