Machine Learning for crash event detection from black-box data This thesis focuses on the statistical analysis of data collected by black boxes in order to detect car crash events. These boxes are installed on private vehicles and are equipped with GPS, accelerometer, and gyroscope. First, unsupervised machine learning and clustering algorithms will be applied to analyse the data and explore possible underlying structures. Then supervised statistical learning techniques will be applied to develop a classifier that is able to predict whether a crash has occurred or not.