Harry Timmermans Eindhoven University of technology Using Intelligent GPS Devices and Learning Algorithms for Collecting Multi-Week Activity-Travel Diary Data: Experiences in a Dutch Context Harry Timmermans Eindhoven University of technology
Introduction Increasingly more complex models set higher demand for travel survey data Activity-based models are gradually moving from static, single-day models to dynamic, multi-week models Conventional activity-travel diaries are insufficient
Basic idea 1. use GPS-device to record information about activity-travel patterns 2. use automated intelligent algorithms to translate GPS traces into activity-travel diaries 3. use web-based prompted recall for validation 4. Use learning algorithms to improve accuracy of imputation over time so that respondent burden will reduce over time and they may be willing to carry GPS device a longer period of time
Evidence
Evidence
Conclusions Findings Future research More difficult study areas Interrupted series due to holidays, day trips etc. Learning algorithm seems successful Imputation results are promising (in this study area) Future research More difficult study areas More difficult segments of the population Using larger set of imputation rules