Inferring the route-use patterns of metro passengers based only on travel-time data within a Bayesian framework using a reversible-jump Markov chain Monte Carlo (MCMC) simulation

• The present study proposed a robust methodology to recognize route-use patterns of transit users.
• The number of used routes was incorporated as an unknown parameter into a Bayesian framework.
• The present study estimated the passenger share and the mean and variance of route travel times.
• The present approach showed a good performance in reproducing the observed number of routes used.

The passenger share and the average travel time for multiple routes connecting an origin–destination pair on a metro network has been examined based on a known number of used routes. Determining how many routes were used based only on travel times from smart-card data is a difficult task, even though the automatic fare collection system can provide a massive amount of travel data. The present study proposes a robust approach to incorporate the number of used routes as an unknown parameter into a Bayesian framework based on a reversible-jump Markov chain Monte Carlo (MCMC) algorithm. Other route-use patterns such as the passenger share and the mean and variance of route travel times were also estimated. The performance of the present approach was compared with the existing method, which depends on the Bayesian information criterion (BIC). The present approach showed better performance in reproducing the observed number of routes used, and also provided greater flexibility in recognizing route-use patterns through the marginal posterior distribution of other unknown parameters.