No open-access location is known for this record yet.
No open-access location is known for this record yet.
As a famous step in the transportation planning process, the mode choice problem has been traditionally studied by employing the concept of random utility optimization in the last 50 years. More recently and especially in the past decade, the emergence of big data and the computational power and insights brought by machine learning techniques has resulted in incorporating data-driven approaches that can lead to more accurate predictions. In this study, we developed travel mode choice models for commute trips in Chicago. Our data set is obtained from the Chicago Metropolitan Agency for Planning (CMAP) travel tracker survey conducted in 2018–2019, including approximately 12,000 households’ activity and travel records for 24 h. We also employed Google Maps Direction API to calculate travel times. We first constructed the home-based tour of each person’s trips to detect the main choice and eliminate the redundant non-choice trips. Then, we estimated a multinomial logit model to characterize the underlying factors that affect the decision of people in selecting commute travel mode. As one of our contributions and for the first time in Chicago, we have considered seven travel modes, including walking, biking, walking to transit, driving to transit (i.e., park and ride), auto driver, auto passenger, and TNC in the choice set. We also analyze the value of time to draw a comparison between auto, transit, and TNC travel modes. Moreover, we employed machine learning classifiers to model the mode choice problem and compare the performance of these models with the econometric model. Also, a feature importance analysis is performed to further scrutinize the most determining factors of the model prediction. The results of this study can improve the existing mode choice model of Chicago through the inclusion of an emerging mode (i.e., TNC) and better representation of existing modes using supply and operational characteristics.
Loading…
Stored only when both papers are in the corpus — 5 citations exist upstream.
Papers that cite the same references — the relational fallback while embeddings are still backfilling.
Imported from OpenAlex (CC0) — nothing here is generated.