Bicycle Rental Predictive Modeling with Machine Learning for Carbon Neutrality

This research examines the role of public bicycle systems in reducing carbon emissions, which significantly contribute to global warming. Centered in Seoul, South Korea-a city with some of the highest carbon dioxide emissions among global metropolises-this study seeks to improve the city's shared bicycle services to address environmental challenges. The analysis is based on five primary datasets. Data on public bicycle usage, new subscriber data, master information data, bicycle road data, and weather data. These datasets identify patterns in user behavior, particularly across various age demographics, and the relationship between bicycle infrastructure and rental volume. The findings highlight that better access to bicycle lanes and more rental stations are crucial in encouraging bicycle use. Additionally, a random forest (RF)-based predictive model is constructed to estimate bicycle rental volume using weather data, achieving a mean absolute error (MAE) of 5.21. The outcomes suggest that strategically enhancing bicycle infrastructure and implementing predictive rental services can significantly boost public bicycle usage, thereby helping to lower urban carbon emissions. Further researches are recommended to broaden data collection across diverse geographical areas and refine predictive models to enhance the effectiveness of public bicycle systems.