Application of machine learning to traffic noise prediction : a case study of Tampere and Oulu

Abstract

Although artificial intelligence technology has experienced considerable growth recently, some fields still have unexplored potential. This thesis investigates the application of machine learning techniques to map traffic noise levels using open data, specifically targeting sub-urban areas. The thesis reviews state-of-the-art machine learning approaches in urban computing, with an emphasis on noise prediction methodologies. The work includes comprehensive feature engineering and examines the potential for transfer learning between two Finnish cities. Finally, an open-data-based noise prediction model is developed, and its performance is assessed across various configurations. While the model's generalization presents challenges, the open data and methodological framework established in this thesis create a strong foundation for future research and applications in noise prediction.