使用分位数回归量化公交客运量预测的不确定性.pdf

1、Quantifying the Uncertainty of Transit Ridership Forecasts Using Quantile RegressionJawad Mahmud HoqueaIan ZhangaGreg ErhardtaDave SchmittbaDepartment of Civil Engineering,University of KentuckybInsight Transportation Consulting Inc.Background:Transit Ridership Forecast AccuracyConference on Innovat

2、ions in Travel Analysis and Planning!=ForecastForecast 100Transit Ridership is about 25%lower than forecast on average,with average deviation of 31%164 Large-Scale transit projects in the US136 of them open with counts availableCompared forecast ridership toobserved ridership two yearsafter opening.

3、Ridership Forecast Accuracy Over the YearsConference on Innovations in Travel Analysis and PlanningWhat Makes A Forecast“Good”?Conference on Innovations in Travel Analysis and PlanningBeneficial to the decision-making Sensitive to the policies Produce metrics useful to decision making Whether the de

4、cision would change for a different forecast and would the unselected decision lead to a better outcome.Correspondence of the forecasts to the observations?Planning in the Face of UncertaintyConference on Innovations in Travel Analysis and PlanningSource:Anam,S.,Miller,J.S.and Amanin,J.W.,2020.Manag

5、ing traffic forecast uncertainty.ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems,Part A:Civil Engineering,6(2),p.04020009.Accuracy and UncertaintyConference on Innovations in Travel Analysis and PlanningAccuracyCloseness of observation and measurement or estimateRetrospective evalua

6、tion of forecast qualityComparison of actual demand and forecasted demandUncertaintyEstimate of the accuracy.Range in which the real value liesProspective modification of forecasts to ensure quality and reliability Range of values possible for actual demandQuantifying Uncertainty in Forecasts Sensit