Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models
| Main Author: | |
|---|---|
| Language: | English |
| Published: |
The Ohio State University / OhioLINK
2019
|
| Subjects: | |
| Online Access: | http://rave.ohiolink.edu/etdc/view?acc_num=osu1546567129433079 |
| id |
ndltd-OhioLink-oai-etd.ohiolink.edu-osu1546567129433079 |
|---|---|
| record_format |
oai_dc |
| collection |
NDLTD |
| language |
English |
| sources |
NDLTD |
| topic |
Civil Engineering Transportation forecast uncertainty point estimates transportation flow models |
| spellingShingle |
Civil Engineering Transportation forecast uncertainty point estimates transportation flow models Bicici, Serkan Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| author |
Bicici, Serkan |
| author_facet |
Bicici, Serkan |
| author_sort |
Bicici, Serkan |
| title |
Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| title_short |
Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| title_full |
Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| title_fullStr |
Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| title_full_unstemmed |
Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models |
| title_sort |
incorporating uncertainty with transportation point forecasts: applications to roadway network and transit passenger origin-destination flow models |
| publisher |
The Ohio State University / OhioLINK |
| publishDate |
2019 |
| url |
http://rave.ohiolink.edu/etdc/view?acc_num=osu1546567129433079 |
| work_keys_str_mv |
AT biciciserkan incorporatinguncertaintywithtransportationpointforecastsapplicationstoroadwaynetworkandtransitpassengerorigindestinationflowmodels |
| _version_ |
1719454971753136128 |
| spelling |
ndltd-OhioLink-oai-etd.ohiolink.edu-osu15465671294330792021-08-03T07:09:26Z Incorporating Uncertainty with Transportation Point Forecasts: Applications to Roadway Network and Transit Passenger Origin-Destination Flow Models Bicici, Serkan Civil Engineering Transportation forecast uncertainty point estimates transportation flow models Travel demand models predict or estimate flows with point values that are used to support planning studies and policy decisions. These model predictions or estimates are subject to multiple sources of errors. Traditional approaches that attempt to incorporate uncertainty in model forecasts usually ignore the errors resulting from model structure and focus on only one or a few sources of errors. We propose an approach that addresses the multiple sources of error, including erroneous model structures. Specifically, we propose “superimposing” error distributions onto the model point outputs to produce cumulative probability distribution functions of forecasts, where the errors are determined from discrepancies between previous model point outputs and observations of the predicted flow variables. We also propose an evaluation approach to assess the performance of the methodology. The evaluation approach is based on determining cumulative probability distribution functions of forecasts using the proposed approach and using corresponding true realizations to determine the cumulative probability values associated with the predictions. These cumulative probability values should be uniformly distributed if the forecasted cumulative distribution functions accurately depict the uncertainty in the true flow variables. Several metrics are developed to measure deviation of these cumulative probability values from the uniform distribution.Investigations based on Metropolitan Planning Organization (MPO) roadway network model and other travel demand model predictions of link flows are presented to illustrate and evaluate the approach under different conditions. Empirical datasets were received from several sources. Using these empirical datasets, we evaluate additive and multiplicative error specifications and see that the multiplicative specification captures the uncertainty better than the additive specification. We also evaluate the benefit of segmenting links by functional class in calibrating error distributions and forecasting. Improved results are obtained when using functional class segmentation. Empirical investigations of the “transferability” of calibrated error distributions across different datasets are also conducted. When forecasting link flows using MPO-based models, better performance is obtained when using network-based MPO and state-wide link flow data to calibrate the error functions than when using project evaluation flow data. However, degradations obtained when using any of these datasets to calibrate the error functions are small compared to degradations obtained when using a naive model that is proposed as a benchmark. Similar investigations are conducted using models to estimate bus passenger origin-to-destination (OD) flows matrices. These investigations are conducted using empirical datasets and using numerical (simulated) datasets. Similar observations to those of the roadway network flows one are also obtained in this application.The empirical and numerical investigations conducted indicate sufficient heterogeneity among the results that if data are available it would be better to use data from the same agency for the same application to calibrate error functions to be used when incorporating uncertainty in forecasts based on model point predictions. However, sufficient data do not appear to be available for general use at this time, and degradations obtained when using heterogeneous datasets are seen to be relatively small when compared to degradations obtained from benchmark naive models. Therefore, default error distributions are developed. 2019-08-28 English text The Ohio State University / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=osu1546567129433079 http://rave.ohiolink.edu/etdc/view?acc_num=osu1546567129433079 restricted--full text unavailable until 2023-05-06 This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws. |