应用K2算法和贝叶斯参数估计方法,进行了贝叶斯网络的结构和参数学习,建立了停车行为分析的贝叶斯网络。应用连接树传播算法推断停车费率影响下的停车开始时间、停车时长和停车场类型等选择行为的变化,预测停车收费政策的实施效果,评价政策的可行性.结果表明：随着停车费率的提高,停车者更趋向于选择短时间停车;对不同时段和不同停车场类型实施不均衡收费制度,即高峰停车费率大于非高峰停车费率,路内停车费率大于路外停车费率,可以促使停车者选择非高峰时段停车和路外停车.

In this paper,first,a Bayesian network for parking behavior analysis is constructed based on the structure and parameter learning by means of K2 algorithm and Bayesian evaluation. Then,with the help of junction-tree propagation algorithm,the variations of parking starting time,parking duration and parking lot type with the parking fee are analyzed. Finally,the effect of parking pricing policy is predicted and the feasibility of the policy is evaluated. The results indicate that high parking fee may shorten the parking duration,that a disequilibrium pricing policy,that is,high parking fee for peak-period parking and on-road parking,may result in non-peak and off-road parking.