出租车乘车概率预测的置信规则库推理方法

Belief Rule-Base Inference Methodology for Predicting Probability of Taking Taxi

出租车乘车概率预测中存在数据量级大,底层属性类型多,预测信息不确定的问题。鉴于此,整合大规模轨迹数据范畴中现有的挖掘算法对出租车GPS数据和路网数据进行离线处理;将多类型的不确定性数据转换为具有置信结构的规则形式,并以此构建置信规则库;通过置信规则库推理方法(belief rule-base inference methodology using evidential reasoning,RIMER)在线预测路网道路上各个地点的乘车概率。以北京市2012年11月某天的出租车GPS数据为例说明该在线预测方法的应用。实验结果表明,该预测方法具有较高的实时性和准确性。

Large scale of data, various types of low-level attributes and uncertainty of prediction information exist in probability prediction of taking taxi. To solve these problems, this paper offline deals with the GPS data of taxi and road network data by using mining algorithms in the large-scale trajectory data domain, then builds a belief rule-base by transforming various types of information with uncertainty into rules which are in form of the belief structure, after that uses RIMER （belief rule-base inference methodology using evidential reasoning） to get the final probability of any points on the road network. Finally, the GPS data of Beijing＇ s taxi in November of 2012 are taken as an example to illustrate the usage of the online prediction method, and the results show the real-time and accuracy of the proposed method.