Data-driven prediction of harsh driving events in urban mobility: an interpretable ensemble approach

I am pursuing a master’s degree at Erasmus University Rotterdam (Data Science & Marketing Analytics).

Timeline: I would like the research proposal to be ready by 2 September.

I would like your help in writing the research proposal and, subsequently, the thesis

I will implement the data analysis and the code myself; the thesis will be in APA style, approximately 50 pages.

There is a repository with samples of the same program for reference. I am doing an internship at Bridgestone and will use corporate telematics data. At a high level, the analysis is based on GPS traces and driving events (speed, direction, speed limits, road type, consumption, etc.). The goal is a quantitative predictive classification of “harsh events” (e.g. sudden braking/turning). I will start with Logistic Regression (baseline) and compare with Random Forest and XGBoost, evaluating performance and interpretability of features.