Adaptive parking demand prediction using discrete time based dynamic Markov chain

The demand for urban parking rapidly increases and becomes a significant traffic issue in densely populated metropolitan regions. Prediction of parking demand is crucial for reducing traffic jams and decreasing greenhouse gas emissions. It is also essential to the development of parking facilities and price adjustments in urban parking planning. Most of the earlier studies developed model for parking demand prediction using historical data which lack to update the demand data. Furthermore, the demand predictions are not considering the effect of parking pricing. However, parking pricing affects the demand in a given parking platform. To address this issue, we have considered three categories of parking demand based on price based preference. Dynamic non-homogeneous Markov chain with discrete time and discrete state is used to predict the parking demand. An adaptive approach or a learning approach is proposed to make the Markov chain dynamic and to adapt changes in the demand environment. A numerical example demonstrating the prediction from data collection as well as incorporating the adaptive strategy so that the system learning new changes, is presented.