This study explores the potentials of utilizing GPS data for Activity-Based Models, under the condition that no additional information, such as travel diaries, is required. To extract activity details, we first developed a Time-Spatial Centroid-based Clustering algorithm to identify activity locations and times. Then a Home Detection algorithm was used in combination with two APIs from Google Maps, namely Nearby Search and Place Details, to label each identified activity as either ‘home’ or a specific activity type. Next, a Markov Chain Multinomial Logit Choice model was developed for the extracted activities that models the sequential relationship between consecutive activities. The approach was applied to a GPS dataset collected in Japan in 2020. The estimated parameters revealed how background factors, such as activity time and the person's age, and the previous activity associate with the current activity. Thus, GPS data alone can provide certain knowledge about activity-travel, which potentially benefit practices of travel demands forecasting.
Activity-Based Model using GPS Data and Google APIs
08.10.2022
517966 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Empirical MFDs using Google Traffic Data
IEEE | 2018
|Online Contents | 2007
|Online Contents | 2006
|