Department of Computer Science: MSc Thesis Presentations

Weather-Based Prediction of Vehicle Malfunctions in Finland

Author: Roope Räsänen
Supervisor: Juho Rousu
Time: Thursday 19 December at 14:30-15:00
Place: online (zoom) https://aalto.zoom.us/j/66875473808

Abstract: This thesis investigates the relationship between weather conditions and vehicle malfunctions in Finland, developing predictive models to enhance the efficiency of vehicle repair services. Through analysis of extensive datasets from Autoliitto and the Finnish Meteorological Institute, the study examines how different weather parameters influence vehicle reliability across various regions and time periods. The research implements and evaluates multiple machine learning approaches, including gradient boosting methods and transformer models, for predicting weather[1]related vehicle malfunctions. Results demonstrate that gradient boosting achieves the best performance, with R² values of 0.4702 for all malfunctions and 0.3322 for battery-specific issues. The study reveals significant regional variations in weather sensitivity, with urban areas showing stronger correlations between temperature and malfunction rates compared to coastal regions. Key findings include a 183.1% increase in battery malfunctions during very cold conditions and a 66.5% increase in overall malfunctions. The analysis identifies critical time windows for weather effects, with 3-6 hour periods being most predictive of malfunction occurrences. Regional variations prove crucial, with major urban areas showing distinct patterns compared to rural and coastal regions. The thesis provides practical recommendations for various stakeholders, including vehicle manufacturers, service providers, fleet operators, and individual vehicle owners. These insights enable more efficient resource allocation and improved preventive maintenance strategies, particularly in regions with extreme weather variations. The findings contribute to the understanding of weather-related vehicle malfunctions and establish a foundation for enhanced predictive maintenance services in cold climate regions.