Understanding parking decisions with a Bayesian network

In this paper, a Bayesian network is developed to investigate three intertwining parking decisions, namely parking period, parking location, and parking duration, and the impacts of a number of parking-related factors on these decisions. With parking information from Beijing, China in 2005, the structure and parameter of a Bayesian network were learnt by employing the K2 algorithm and Bayesian parameter estimation method respectively. The results show that the decision on how long to park follows that on where to park, and both of them are affected by the decision of when to park. This suggests that parking policies aimed at intervening in one specific parking decision may have an indirect influence on other parking decisions, which embraces an integrated view in the development of parking policies. The findings facilitate the development of measures for regulating parking behavior by identifying important contributing factors.