Navigating Zurich: A Comprehensive Analysis of Urban Traffic Dynamics

Pascal Sager*, Luca Zhao*, Weijia Zhong*, Xiaohan Zhu*
Zurich University
Project Work: Introduction to Data Science
*Equal Contribution
Project Proposal Code Presentation

Abstract

Public transportation systems play a crucial role in urban mobility, and analyzing their dynamics is essential for optimizing efficiency and planning for future needs. In this study, we employ advanced data analytics techniques to delve into the public transportation system of Zurich, Switzerland. Leveraging publicly available CSV datasets released by the Canton of Zurich, we explore the temporal evolution of transportation metrics, including spatial coverage, passenger volume, and utilization intensity. Additionally, we investigate the spatiotemporal distribution of passengers relative to vehicle capacity, examining factors such as spatial location, weekday patterns, and academic calendar influences. Our analysis is guided by fundamental research questions encompassing historical trends, comparative analyses with other global cities, and predictive modeling of seat availability. While our findings offer valuable insights into Zurich's transportation landscape, we acknowledge the inherent limitations stemming from dynamic factors such as evolving traffic schedules and political decisions. Nonetheless, our study provides a comprehensive framework for understanding and potentially enhancing public transportation systems in Zurich and beyond.

Overview Results

Below, we present some results in an overview. For more detailed results, please refer to the respective sections and have a look at our presentation.

Development of Zurich’s Public Transportation System Over Time

We analyzed the evolution of Zurich's public transportation system by computing the total distance coverage and passenger volume for each year from 2014 to 2023. Our analysis revealed that the total distance coverage remained relatively stable over this period. Passenger volume also exhibited a consistent trend, except for a significant decrease in 2020 and 2021, likely due to the impact of the COVID-19 pandemic.

Further investigation was conducted to understand the contributions of different modes of transportation to these metrics, specifically focusing on trams, buses, and trolleybuses. We also included the Forchbahn, Seilbahn, and the night network in our analysis.

An interesting observation from 2022 to 2023 was a noticeable shift in passenger volume from the bus system to the tram system. This shift indicates a potential change in passenger preferences or operational efficiencies within the public transportation network.

Overall, our study highlights the resilience of Zurich's public transportation system in maintaining service coverage and adapting to changing passenger demands.

Development of Zurich’s Public Transportation System Over Time

We hypothesized that the average number of passengers using Zurich’s public transportation system is affected by factors such as weekday/weekend and lecture period/exam session, particularly at stops in the vicinity of universities.

Therefore, we computed the average number of passengers after grouping the data accordingly. Interestingly, we observed more passengers on average on weekdays, likely due to commuting needs.


Occupation per Daytype

Comparing exam days to the lecture period, it was surprising to see a decreased passenger number during exam periods, regardless of proximity to universities. Specifically, at the ETH/Hönggerberg stop, which is predominantly used by students, passenger behavior did not differ significantly. Seasonal factors, such as extreme weather conditions during exam sessions, might have a significant confounding effect.


Influence Exam Sessions

Links to Traffic Maps

Time Weekday Vehicle Occupation Weekend Vehicle Occupation Weekday Passengers/Stop Weekend Passengers/Stop
0.00am - 1.00am Map Map Map Map
1.00am - 2.00am Map Map Map Map
2.00am - 3.00am Map Map Map Map
3.00am - 4.00am Map Map Map Map
4.00am - 5.00am Map Map Map Map
5.00am - 6.00am Map Map Map Map
6.00am - 7.00am Map Map Map Map
7.00am - 8.00am Map Map Map Map
8.00am - 9.00am Map Map Map Map
9.00am - 10.00am Map Map Map Map
10.00am - 11.00am Map Map Map Map
11.00am - 12.00am Map Map Map Map
12.00pm - 1.00pm Map Map Map Map
1.00pm - 2.00pm Map Map Map Map
2.00pm - 3.00pm Map Map Map Map
3.00pm - 4.00pm Map Map Map Map
4.00pm - 5.00pm Map Map Map Map
5.00pm - 6.00pm Map Map Map Map
6.00pm - 7.00pm Map Map Map Map
7.00pm - 8.00pm Map Map Map Map
8.00pm - 9.00pm Map Map Map Map
9.00pm - 10.00pm Map Map Map Map
10.00pm - 11.00pm Map Map Map Map
11.00pm - 0.00am Map Map Map Map

Predicting Seat Occupancy

We use various techniques to predict the seat occupancy of vehicles in Zurich. We train the model on data from 2022 and test it on data from 2023 to estimate how well the model generalizes for future predictions. As a baseline, we use the median seat occupation. We first show, that simple models such as linear regression and random forest regressors do not work well and perform similar than the baseline. To use support vector classifiers, we first reduce the dimensionality of the data using UMAP as otherwise, given our hardware, training is not feasible. We then show that the support vector classifier performs better than the baseline and the other models. Finally, we show that deep networks perform even better than the support vector classifier. For more details on the implementation, please have a look at the readme file and the notebooks in the code repository.

Furthermore, we conduct various analyses to understand the model's behavior and the data. We show that adding geospatial data, that we obtain from external sources, drastically improves the model's performance.

BibTeX


@misc{sager_zhao_zhong_zhu_2024,
  author       = "Sager, Pascal and Zhao, Luca and Zhong, Weijia and Zhu, Xiaohan",
  title        = "Navigating Zurich: A Comprehensive Analysis of Urban Traffic Dynamics",
  month        = "May",
  year         = "2024",
  note         = "Zurich University Project Work: Introduction to Data Science"
}