Publication

Abstract : The recent years have seen a rise in ride-hailing services. As a result, improving surge pricing techniques has become a crucial challenge where both customer satisfaction and company revenue are to be balanced. Surge price prediction can be implemented using real-time data and machine learning models. Ride fares change based on numerous factors such as demand for rides, time of day, traffic, and driver availability. Taking all these variables into account poses a huge challenge for surge price prediction. Therefore, the study deals with complicated, imbalanced ride-hailing records by using an ensemble of Random Forest and XGBoost with SMOTE preprocessing to specifically predict sudden pricing types. Each model is trained and tested individually, where Random Forest achieves an accuracy of 86%, while XGBoost achieves an accuracy of 89%. The ensemble model increases the accuracy to around 93%, demonstrating superior predictive performance compared to the individual models. The proposed approach helps companies optimize revenue, improve driver distribution, and dynamically adjust pricing during peak demand while responding effectively to market changes and customer expectations.