Choice set generation from GPS data set for grocery shopping location choice modelling in canton Zurich: Comparison with the Swiss Microcensus 2005

Abstract

This paper describes a set of new procedures used to generate choice sets for grocery shop-ping destination choices observed in a large scale GPS data set. The procedures are a rule-based identification of the shopping trips, defining places from the observed locations using a hierarchical-clustering algorithm and finally a rule-based identification of the place of home and work/education and implicitly with this the imputation of the trip purposes.<br/>The GPS traces were collected by volunteers: 2,435 persons in Zurich and 1,086 persons in the smaller city of Winterthur. No socio-demographic information such as age, sex, income, etc. was available for this research, as the data was collected for a different purpose. The choice sets are generated by transport mode used for the shopping trip; car, walking and urban public transport whereas trips by intercity rail and bicycle were excluded since their numbers are too low for an estimation of a destination choice model, the ultimate purpose of the overall project.<br/> The results were compared with the Swiss Microcensus (MZ) 2005, the national travel behav-iour survey, for time-of-week, time-of-day, duration, duration of the shopping activity. While the quality of the results was high, the comparisons highlight instructive differences resulting from the different data generation process.