基于贝叶斯方法的出行产生率预测

Bayesian Updating of Trip Generation Rate

基于贝叶斯统计方法的理论分析,结合贝叶斯方法在交通参数预测中的既有应用,提出了贝叶斯方法预测国内出行产生率的具体方法.该方法以大规模居民调查历史数据为出行产生率的先验分布,以最近小样本抽样数据作为新样本数据,近期出行产生率则为需要预测的后验分布期望值,在先验样本与新样本均超过30的前提下,将先验样本与新样本的均值与方差视为正态分布无偏估计,进而应用贝叶斯统计方法计算得到近期出行产生率,从而达到小样本抽样调查即可对出行产生率进行及时更新的目的.以实例验证了该方法的可行性,并指出其局限性与适用条件.

This paper introduces a specific method to update short-term trip generation rate applicable in Chinese cities on the basis of theoretical analysis of Bayesian statistical methodology and previous method of updating parameters related to transportation analysis. According to this method, the historic trip investigation data are assumed to be the prior information and prior distribution, a limited number of recent cases are taken as new sample, and the short-term trip generation rate taken as posterior probability to be predited. If the number of historic and new samples are both more than 30,and their mean and standard deviation are assurned to be the unbiased estimator of the prior and sample variance, the expectancy of posterior distribution, viz. the short-term trip generation rate, can be calculated using Bayesian statistical methodology. The feasibility of this method is exemplified and its deficiency and suitability are also pointed out.