Cycling behavior in Germany - a multi-criteria approach to quantify local differences in route choice and driving behavior.

• Applicants: Dr.-Ing. Stefan Huber; Dr.-Ing. Sven Lissner; Dr. Iryna Okhrin
• Subject Area: Urbanism, Spatial Planning, Transportation and Infrastructure Planning, Landscape Planning
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 514587991

Cycling has increasingly gained attention in traffic research in the past decade. Research and qualification have mostly been based on data from classical household surveys (e.g., MiD, SrV) or movement data for single cities. However, household data do not provide any information about cycling behavior or route choice, and movement data of single cities cannot produce general statements about how cycling depends on city characteristics.The research project “Cycling behavior in Germany” aims to analyze the influence of spatial and city-specific variables on cycling behavior and route choice in Germany. The goal is to answer the question, for example, why people in Münster have a different cycling behavior than people in Berlin and why other parameters may dominate for the route choice in Dresden than in Stuttgart. The project’s results should thus ensure a basic transferability of route choice models to different municipalities.The available dataset contains more than 3 million GPS trajectories of about 250,000 cyclists in Germany. The analysis will be based on carefully selected cities. They have to be as diverse as possible in terms of their characteristics (e.g., size, topography, the proportion of cycling in the modal split) and, at the same time, represent best all the German cities. The detailed information about the selected cities will be merged with the secondary data from GPS trajectories. Thus an extensive dataset will be created, which allows for a comparative analysis of cycling behavior and route choice in the studied areas. In order to analyze the data set both established procedures and new methods are developed and applied for the statistical analysis. In addition to classic econometric methods (e.g. multinomial logistic regression), a new methodological approach will be developed that includes different machine learning (ML) methods, which will be used for the analysis of cycling and route choice behavior. To compare the results of the analysis across different models (econometric and ML), a separate methodology will be developed as part of the project, too.For the analysis of cycling behavior and route choice, well-known methods are planned to use used as well as new methods will be developed and applied. Besides classical statistical methods (e.g., multinomial logistic regression), different machine learning (ML) methods will be applied. For the comparison of cycling behavior and route choice across different models (statistical and ML), the development of an own methodology is planned in the scope of this project.As a result, global and local variables and their strength in the cycling behavior and route choice in Germany will be identified and quantified. The use of recent ML methods and their comparison with each other as well as with statistical methods intersects with traffic econometrics and statistics as well as with recent trends in the field of artificial intelligence.

• DFG Programme: Research Grants