No abstract

GPS tracking technology produces large amounts of data which represent samples of the commercial vehicle population that are much larger than conventional commercial travel surveys. However, passively collected GPS data lack behavioral detail that a conventional survey offers. This study develops a data fusion method to impute variables of interest for a large GPS data set, by establishing a link to a behaviorally rich commercial travel survey data set. As a case study, this study uses detailed information from the Ministry of Transportation of Ontario’s Commercial Vehicle Survey (CVS), a truck intercept survey conducted in 2012, to enrich a GPS commercial vehicle tracking data set from Xata Turnpike Inc. The enrichment process has three parts: converting raw GPS tracking data into GPS trips, matching CVS trips to GPS trips, and imputing the missing variables for GPS trips. Evaluation of the outcomes concludes that imputation methods can produce a synthetic data set with large sample size (from GPS data) and rich information (from roadside interview data) with good accuracy.