University of Twente Student Theses

Login

Discrete Choice Modeling for Transfer Flow and Partial Multimodal Public Transportation OD Matrix Estimation

Lang, G.S. (2024) Discrete Choice Modeling for Transfer Flow and Partial Multimodal Public Transportation OD Matrix Estimation.

[img] PDF
16MB
Abstract:Public transportation is a key component of sustainable urban mobility and ensuring its efficiency and quality of service is pivotal for its success. To this end, analyzing travel demand and passenger flow through the system is an important task for public transportation providers, as they can identify where bottlenecks arise or the most important connections and propose countermeasures. The network-level Origin-Destination (OD) matrix is the primary tool that is used to enable this analysis, as it captures the origin and destination stop of a passenger, as well as the lines taken to arrive there. However, the process of estimating the network-level OD matrix is time-consuming, costly, and heavily data-reliant, with only the technological and methodological advancements in recent years allowing for effective and accurate estimation. This advancement in methodology came with the arrival of automatic data collection systems, such as automatic fare collection systems or GPS, which allowed public transportation providers to more effectively track individuals throughout their network. On the other hand, though these systems have become more widespread, not all of them are present in every public transportation system, and the estimation of the network-level OD matrix remains complex. In particular, one of the harder patterns to capture is the transfer flow at a station, where travel surveys, due to sample size, often do not capture the whole picture, leading to many structural zeros. Two methodologies are proposed using Multinomial Logit Models, a travel sample, network data, boarding and alighting counts, and schedule data, with which the transfer flow patterns at a station may be estimated and analyzed, and subsequently utilized to construct the network-level OD matrix. The choice set of method 1 consists of all lines passing at the transfer station, to which a passenger would reasonably transfer, as well as exiting the network, indicating this is their final destination. The choice set of method 2 is then an expansion of that of method 1, where each line is disaggregated into the zones they stop in. To describe the transfer patterns, sets of 9 and 13 explanatory variables are defined, for methods 1 and 2, respectively, from which the tractable and statistically significant ones were singled out. These methods were applied to a chosen transfer station of the Geneva public transportation network, Lancy-Bachet. The results of this application show that the model fit is ρ¯ 2 ≥ 0.4 for all models, which indicates that both methods can effectively explain the transfer and transfer-destination flow. Furthermore, it is shown that method 2 outperforms method 1, as it is less prone to errors. In the case of both methods, the explanatory variables with the highest coefficient dominance and contribution to model fit are the Intercept, Transfer time, and Angle cost, with the addition of Travel time for method 2. However, although the overall fit is shown to be excellent, both methods are subject to wrongly estimating low flows (< 10 passengers/hr), where errors are perceivable. Furthermore, though method 1 is shown to apply to Lancy-Bachet, it is found that at other stations of the Geneva network, the coefficient values lose interpretability. This behavior is not shared by method 2, where the calibrated coefficients are tractable. It is thus concluded that, although the data requirement is higher, method 2 is more suitable for practical use in estimating travel demand and transfer flow, than method 1. In addition, it is found to more accurately explain transfer and travel behavior, and is recommended to be incorporated in network-level OD matrix estimation as required.
Item Type:Essay (Master)
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering and Management MSc (60026)
Link to this item:https://purl.utwente.nl/essays/102591
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page