The Application of XGBoost and SHAP to Examine Factors in Bike Sharing-Related Demand

Due to the burgeoning demand for bike sharing usage, bike sharing related urban management issues have been growing substantially. This article used data from the Beijing region in 2019 to explore the relationships between the weather factors and cycling demand, and predicted the demand of bike sharing using XGBoost. Then, SHAP was employed to interpret the results and analyze the importance of individual features. The results showed that the overall accuracy of the XGBoost method (82%) was higher than the multiple ordered logistic regression model (72%). The important influencing factors, revealed by two models, included temperature, humidity, power of the wind, and PM10. The study found that when the temperature is lower than 20 degrees Celsius, the humidity is higher than 25, the power of the wind is greater than level 7, or the concentration of inhalable particulate matter is too high, the cycling demand will be significantly reduced. Therefore, it is necessary to set the number of shared bicycles available according to different weather conditions. Cycling infrastructure should also be improved to ensure the usage of bike sharing under adverse weather conditions. This article also provides decision support for the sustainable and healthy development of bike sharing.