Evaluating car-sharing switching rates from traditional transport means through logit models and Random Forest classifiers

ABSTRACT

Positive impacts of car-sharing, such as reductions in car ownership, congestion, vehicle-miles-traveled and greenhouse gas emissions, have been extensively analyzed. However, these benefits are not fully effective if car-sharing subtracts travel demand from existing sustainable modes. This paper evaluates substitution rates of car-sharing against private cars and public transport using a Random Forest classifier and Binomial Logit model. The models were calibrated and validated using a stated-preference travel survey and applied to a revealed-preference survey, both administered to a representative sample of the population living in Turin (Italy). Results of the two models show that the predictive power of both models is comparable, albeit the Logit model tends to estimate predictions with a higher reliability and the Random Forest model produces higher positive switches towards car-sharing. However, results from both models suggest that the substitution rate of private cars is, on average, almost five times that of public transport.

KEYWORDS:

• Travel demand
• sustainability
• car-sharing
• data mining
• mode choice
• multimodality
• case study

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was partly sponsored by the European Union-funded project ‘Shared mobility opporTunities And challenges foR European citieS’ (STARS), which has received funding from the European Union's Horizon 2020 research and innovation program under Grant Agreement no. 769513. Financial support to collect the SP dataset also came through a ‘Ricerca dei Talenti’ grant from Fondazione CRT.