Machine learning-based analysis of environmental impact on cycling behavior: A study across multiple Nordic cities

Abstract

As global urbanization accelerates, traffic congestion and environmental pollution have become critical issues. Cycling, as a green transportation mode, is essential for promoting sustainable urban development. However, the impact of environmental factors on cycling behavior varies among cities, and systematic comparative studies are lacking. This study focuses on five Nordic cities—Reykjavik, Oslo, Stockholm, Copenhagen, and Helsinki—utilizing multi-source data and an improved NRBO-XGBoost model to enhance predictive performance in analyzing the effects of environmental factors on cycling mobility. Through K-means clustering, differences in cycling mobility patterns among these cities are revealed. The results indicate that topographical variation, green space area, air quality, and road density significantly influence cycling flow, with varying impacts across cities. In Copenhagen and Oslo, high densities of commercial facilities and green space coverage promote cycling activity, whereas complex terrain in Stockholm and Helsinki restricts it. Cluster analysis shows that high traffic network density in the city centers of Oslo and Copenhagen enhances cycling flow, and suburban green spaces and water bodies contribute to increased cycling. Conversely, slopes in Stockholm and Helsinki inhibit cycling activity. These findings equip urban planners with both generalizable and targeted insights to optimize cycling infrastructure and promote sustainable urban mobility tailored to each city’s unique environmental conditions.