Stochastic modeling and adaptive forecasting for parking space availability with drivers’ time-varying arrival/departure behavior

Parking space availability is valuable information to travelers. This paper aims at modeling drivers’ behavioral changes in arrivals/departures over time of day and developing an adaptive forecasting approach for parking space availability. We propose a stochastic model that consists of two inter-connected Markov processes. First, the lower level of the model focuses on the parking behavior within a short time period, based on conventional M/M/C/C queueing theory with the assumption of fixed arrival and parking rates. Next, to account for the behavioral changes in drivers’ arrivals/departures over a longer time period (e.g. time of day), we incorporate a Markov regime switching process to describe the regime switching mechanism of the arrival/departure behavior. The integrated model leads to an adaptive forecasting formula with time-varying forecasting coefficients adaptively adjusted based on the arrival/departure regimes. We investigate two real traffic applications to illustrate the developed stochastic model and to test the performance of the adaptive forecasting method using out-of-sample data.