Utilizing P-median and Machine Learning for Bike-Sharing Demand Prediction

Bike-sharing systems (BSS) are under careful examination for their potential to improve urban accessibility and sustainable transportation. Despite their widespread adoption, effectively managing BSS encounters difficulties arising from imbalances between demand and supply. The key challenge in BSS operation is accurately forecasting bike demand across different stations. The task was addressed in this paper through regression techniques such as Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Regularized Linear Regression (Ridge), and Least Absolute Shrinkage and Selection Operator (LASSO) for short-term predictions. n this study, we utilize comprehensive datasets encompassing bicycle-sharing activities and weather conditions in Los Angeles. Integrating these datasets with the p-median method for station grouping, we examine the delicate balance between median values and prediction accuracy. Furthermore, our investigation reveals that among the various models scrutinized, RF and XGBoost consistently exhibit superior performance.