This thesis investigates the impact of machine learning on solving classification problems and supporting strategic decisions, particularly within the automotive and fleet management sectors. Conducted in collaboration with VivaDrive, a technology company focused on fleet optimization and electrification, this project applies machine learning to real-world data to guide the transition from conventional to electric vehicles (EVs) and identify the best machine learning algorithms for each classification task.
The thesis focuses on three key classification problems using VivaDrive’s operational data:
-
Binary Classification: Predict whether a conventional vehicle is suitable or unsuitable for replacement by an EV.
-
Multiclass classification - Suitability Scoring: Predict how suitably (ranked categories) a conventional vehicle can be replaced by an EV.
-
Multilabel Classification: Predict all the suitable electric vehicle brands that can replace a given conventional vehicle.
Final versions:
- Thesis in English
- Thesis in Greek, original version.