Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Dieses Kapitel geht den Herausforderungen nach, vor denen Verkehrsplaner stehen, wenn es darum geht, umfassende Herkunfts- und Bestimmungsmatrizen für Fahrten zu erhalten, die durch Bildschirmlinien wie Stadtgrenzen führen. Die Autoren schlagen eine Methode vor, die Floating Phone Data (FPD) und räumliche Cluster-Techniken nutzt, um die Verteilungen der Reiseziele abzuschätzen, und adressieren die Beschränkungen traditioneller Erhebungen mit kleinen Stichprobengrößen. Die Studie stützt sich auf vier Datenquellen: räumliche Strukturdaten, OD-Routing-Informationen, Erhebungsdaten zu Reisezwecken und von FPD abgeleitete OD-Matrizen. Durch Anwendung eines k-Mittel Clusteralgorithmus gruppieren die Autoren räumliche OD-Daten zu sinnvollen Clustern und übertragen dieses Clustern auf die Erhebungs- und FPD-Datensätze. Dieser Ansatz ermöglicht die Berechnung von Trip-Purpose-Verteilungen pro Cluster, die dann auf die FPD-OD-Matrizen abgebildet werden. Die Methodik wurde erfolgreich auf eine Bildschirmlinienanalyse in Graz angewendet, die zu einer detaillierten Verteilung des Ausreiseverwendungszwecks für jeden Querschnitt der Bildlinie führte. Die Ergebnisse waren mit denen traditioneller Umfragen vergleichbar und zeigten die Wirksamkeit der Methode. Das Kapitel schließt mit Empfehlungen zur weiteren Verfeinerung und Validierung der Methode, wobei ihr Potenzial für eine breitere Anwendung in der Verkehrsplanung hervorgehoben wird.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
Origin-destination (OD) matrices from floating phone data (FPD) are a valuable data source for screen line analysis. However, FPD does not contain any trip purpose information. To get additional information regarding the trip purpose distribution at a screen line, manual surveys with a huge sample size would be needed. Such surveys are very complex and expensive. Hence, they are rarely conducted in practice.
This study aims to overcome these shortcomings and presents a novel approach for assigning trip purpose distributions to OD-matrices. Based on limited but geocoded survey data, spatial structural data and routing information trip purpose distributions can be mapped on OD-matrices from readily available FPD. By using k-means based clustering techniques the spatial structure is used as a link between FPD and survey data.
The developed methodology was successfully applied to the Graz region in Austria, with approximately 300.000 residents and showed promising results. The derived trip purpose distributions were verified by utilising traditionally survey data. Therefore, this method is transferable and can be used as supplement to traffic volume screen line analysis to gain valuable planning information while keeping the survey effort and related costs to a minimum.
1 Introduction and Motivation
For the purpose of demand-oriented planning, transport planners require comprehensive origin-destination (OD) matrices for all trips passing through a screen line, such as a specific segment of the city boundary. This requirement can be fulfilled either by a well calibrated demand model or unbiased OD surveys with a suitable sample size. However, many metropolitan areas lack either of these data sources, yet they require such information for transport planning tasks. Utilizing passively generated Floating Phone Data (FPD) can minimize data collection efforts and improve accessibility to the required information. For instance, OD relationships, traffic volumes, and modal splits (the distribution of transport modes used) at the city boundary can be predominantly acquired from passively generated sources like permanent counting stations and FPD [1].
However, the distribution of trip purposes still requires manual collection, since the motivation behind a trip cannot be passively observed. Moreover, even if a traditional survey is conducted, the small sample size often hinders the usability and accuracy of such data. The proposed method aims to overcome this challenge by using multiple data sources and exploiting the known spatial dependencies associated with trip purposes. This dependencies, include among others, the different types of land use in a destination zone [2, 3] as well as the traveled time and distance to reach this zone [4].
Anzeige
A variety of clustering techniques for estimating trip purpose distributions from different input data can be found in the literature. Most researchers use either land use information of traffic zones as trip purpose distributions representations for destination traffic [2] or cluster travel survey data to obtain groups of passengers with similar travel behavior and enrich given OD-data with this information [5]. Most studies in this research field use OD-data for public transport, like smartcard or automated fare collection data [2, 3, 5] or trajectories from GPS supported surveys as an OD data source [6].
Different to these approaches, we use passively created OD-matrices from FPD for this study, which are easily and continuously available on a large scale while reducing the survey effort. The proposed method in this paper uses clustered spatial structural data as link between given OD-matrices and OD-geolocalized travel survey data. Therefore, we use limited structural data and travel distances between all traffic zones to generalize the information from the trip purpose survey data, which has a small sample size, and map the distributions onto the FPD OD-matrices.
The presented method was successfully applied to a screen line analysis at the Graz Region in Austria, which is a city of 292.630 residents. The screen line of interest is shown in Fig. 1. which represents the city boundary (blue), defined by several cross-sections (red). On disaggregated level we observed 20 different road cross-sections, which represent all major roads passing the screen line (Fig. 1).
Fig. 1.
Screen line and monitored cross sections. Background Map:
In this study we utilized four different data sources: For spatial clustering we used spatial structural data (number of working places, educational places and residents), as well as OD-routing information, like travel time and distance, between all traffic zones. The trip purpose information was derived from an OD-geocoded trip purpose survey which was conducted at 13 out of 20 different cross sections at the screen line. The survey was conducted between 6 am and 6 pm with a sample size of n = 2377 persons. As OD-information we used OD-matrices from FPD which were mapped to the road network. This allowed us to link each trip to a specific cross section, which are shown in Fig. 1.
Anzeige
We monitored only inbound traffic on roads. Furthermore FPD and survey data was weighted with available traffic and occupancy counts similar to [1]. An overview of the used data and developed methodological framework is shown in Fig. 2.
Initially, a spatial OD-dataset was created. This dataset consisted of one instance for each possible OD-relation within the study area. Every OD-relation was identified by the ID of the origin and destination traffic zone and contains data about the number of workplaces, residents and places for educational purpose within both the origin and destination traffic zone, as well as the travel distance and duration between the two zones.
By using a k-means clustering algorithm the spatial OD-data was grouped into meaningful clusters. Before applying the algorithm on the spatial OD-data set the data needed to be standardized, meaning it had to be transformed to have a mean of 0 and a standard deviation of 1. Otherwise this could distort the results of the clustering as data attributes differ in scale [7]. The number of clusters, which is an input parameter for k-means, was determined by calculating the silhouette score. This score indicates the clustering quality based on the previously defined number of cluster [8]. Under the hypotheses that more than two clusters are needed to account for various spatial influences, we identified 9 clusters before observing a significant drop in cluster quality (see Fig. 3).
Fig. 3.
Silhouette analysis for optimal number of clusters.
Each OD-relation in the spatial dataset was assigned to one cluster. The same clusters are transferred to the trip purpose survey and the FPD matrices by matching each OD-relation in both datasets with the corresponding OD-relation in the clustered spatial dataset. The IDs of the origin and destination traffic zones were used as link between the datasets to annotate both datasets with the corresponding cluster.
By utilising the annotated survey data, a trip purpose distribution per cluster was calculated. Each OD-relation in the FPD was assigned to a cluster and the calculated trip purpose distributions from the survey were transferred to the mapped FPD-matrices by use of the cluster annotation. This way we derived a trip purpose distribution for all existing OD-relations in the FPD.
An overview of the described data structure and dependencies is shown in Fig. 4.
As the FPD was mapped to the road network and thus to the cross-sections, the different OD-relations could be aggregated for each cross-section and the screen line, respectively. By aggregating the mapped and annotated OD-matrices a mixed trip purpose distribution was obtained, which represents the final trip purpose distribution at the city limits.
3 Results and Conclusion
The developed methodology is a valuable supplement to estimate trip purpose distributions for OD-data used in screen line analysis. We managed to calculate a detailed trip purpose distribution for each cross-section of the screen line from OD-matrices derived from FPD and manual survey data with a relatively small sample size. Therefore, survey costs were reduced to a minimum. The methodology is transferrable to other regions.
Table 1 shows specific results for our use case in the region of Graz, Austria. The first column shows the aggregated trip purpose distribution for the whole screen line, while the second column shows the upper and lower boundaries of trip purpose distribution at different cross-sections. The last column lists the surveyed trip purpose distribution at the aggregated screen line.
Table 1.
Trip purpose distribution from clustering model compared to survey data.
Trip-purpose
cluster model
(screen line)
cluster model
(cross-sections)
survey data
Work
28.7%
25.0 – 37.7%
31.7%
Education
6.3%
4.9 – 9.2%
5.2%
Leisure
4.6%
3.4 – 5.0%
4.3%
Shopping
7.9%
5.5 – 9.0%
7.3%
Private
14.5%
13.9 – 16.0%
15.0%
Home
29.6%
21.0 – 31.8%
26.4%
Business
3.0%
2.5 – 3.7%
3.0%
Drop off / pickup
5.5%
4.6 – 5.8%
7.1%
The result of the cluster model is comparable and similar in scale to the surveyed distribution. The largest deviation between the modelled results and survey data can be observed for trips back home. While the model shows 29.6% of all trips passing the screen line going back home, the survey only shows 26.4% on this trip purpose. The surveys were conducted between 6 am and 6 pm. Therefore, some late trips back home might be missing in the data, which would be in favour of the higher value for trips back home in the model-based approach. The interval at the different cross-sections shows boundaries from 21.0% to 31.8% of all trips going home, which includes the value of the survey (26.4%). This large dispersion can be explained by different spatial structures along the screen line. Therefore, it’s an expected and desired effect of the cluster model to capture those spatial differences, which impact the trip purpose distributions. For all other trip purposes, the distributions obtained from the model are comparable to the survey data. In most cases the results from the survey data are within the interval of the different cross sections computed by the cluster model. Only for drop off and pick up trips the survey data is outside the interval. This discrepancy could suggest that either the model or the survey fails to accurately capture this specific trip purpose.
A significant benefit of this method is its ability to depict the distribution of trip purposes across all cross-sections with the same level of detail as found in the survey (covering eight distinct trip purposes), despite the constraints of a small sample size. For the method's further refinement, it is advisable to undertake a quantitative validation, on cross-section level, through a comprehensive control survey. For further improvements of the spatial cluster model, we would recommend using additional structural data which can represent the shopping and leisure potential of a traffic zone. These data sources are not always easily available in a good quality. Therefore, this framework and its transferability would benefit from a supplementary method to calculate these potentials from open data.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Lammer, F., Cik, M., Fellendorf, M.: analysis of cross-border traffic for large cities using multiple data sources and new technologies. Transport Research Arena 2022, 15 Nov, Lisbon (2022)
2.
Bo, D.: Estimating Travellers’ Trip Purposes using Public Transport Data and Land Use Information, SMART Infrastructure Facility - Papers (2019)
3.
Alsger, A., Tavassoli, A., Mesbah, M., Ferreira, L., Hickman, M.: Public transport trip purpose inference using smart card fare data. Trans. Res. Part C: Emerging Technol. 87, 123–137 (2018). https://doi.org/10.1016/j.trc.2017.12.016
