Use of passive data for determining link level long distance trips

Abstract

Long-distance trips include a high value of time compared to short-distance trips; thus, capturing long-distance trips contributes to substantial economic and social benefits. This study utilizes privacy-protected travel data collected from mobile devices in Maryland to identify the link-level proportion of long-distance vehicle trips. We propose applying existing econometric frameworks, i.e., the generalized linear and beta regression models, to predict these link-level long-distance trips. Among the covariates, we utilize highway network-level attributes from the Maryland statewide travel demand model (MSTM), employment, and other open-source datasets available for the state of Maryland. Both the econometric models were compared for their performance in multiple measures like validation, behavioral interpretation, and goodness of fit. The best model results indicate a positive relationship between the proportion of long-distance trips and link attributes, like functional class, speed, and surrounding household density. The proposed framework will provide useful insights and key inputs for practitioners in demand modeling in capturing long-distance travel in a roadway network.