Bicycle theft is a huge problem in Berlin, as it has one of the highest rates of bike theft per inhabitant in Germany. This poses the question: How can one keep their bicycle safe in Berlin? Apart from a strong lock and a safe parking spot, there might be other influences on the probability of a bike being stolen, such as place or time of the day. A rather uncommon factor to think of is the weather, though it plays a huge part in mobility and transportation, especially for bicycles. In this project, three datasets from Berlin from 2022 are combined: one containing bicycle theft data, one containing different weather metrics and one containing bicycle traffic counts, looking for anomalies and correlations in the data. The result can be found in the report here.
- RC1: Does more bicycle traffic lead to more bicycle thefts?
- RC2: Are there times where bikes, relative to traffic, are stolen more often?
- RC3: Do higher temperatures lead to more bicycle traffic?
- RC4: Do higher temperatures increase the risk of bicycle theft?
- RC5: Does more rain lead to less or more bicycle traffic?
- RC6: Does more rain increase the risk of bicycle theft?
- RC7: Are there significant differences in weather metrics at times when bikes were stolen?
The /project/pipeline.sh script must be run from the root folder after installing the necessary packages:
pip install -r project/requirements.txt
bash project/pipeline.sh
The /project/tests.sh script must be run from the root folder after installing the necessary packages:
pip install -r project/requirements.txt
bash project/tests.sh
Information about the datasets used can be found in the above mentioned report and in project/project-plan.md.
The data/archived_bicycle_theft_2022.csv is a archived version of the Fahrraddiebstahl in Berlin dataset, which unfortunately is currently no longer available, as it apperently only contains data from the last and current year. It was (and is) published under the CC BY 3.0 DEED license
This project was tested with Python 3.10.
This project is licensed under Creative Commons Attribution 4.0 International.
The banner was generated using DALL-E and cropped.