Files
Comans_72231300_2020.pdf
Open access - Adobe PDF
- 4.32 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Since the introduction of the first motored vehicles to the grand public, road safety has been an important concern for many actors. With the constant increase of new vehicle registrations around the globe and the growing interest in self-driving cars, understanding and preventing traffic accidents is becoming more and more crucial. The objective of this thesis is to apply Machine Learning techniques to model car accidents to help us discover the main factors affecting their severity and creating new metrics measuring the danger of a road. We will base our research on the history of accidents that occurred in Luxembourg between 2013 and 2017 and model them using Decision Trees and Support Vector Machines. We will see that the first can give us a measure of feature importance to consider when studying road characteristics in the context of traffic accident. The second will allow us to estimate a new measure of danger using the calibration method Platt’s scaling in order to represent the risk of having slight, serious or fatal accidents when driving on a certain road. Additionally, we will introduce the concepts of Data Visualization and practical applications when studying mobility data.
Citations
Comans, M.-H. (2020). Modelling crash severity and safe-driving through vehicle dynamics. https://dial-mem.test.bib.ucl.ac.be/handle/2078.2/17538