Development of Machine Learning Models for Predicting Trip Cancellation in Ride-Hailing Platforms.
Year: 2026
Type: article
Abstract: Trip cancellations remain a significant operational challenge in ride-hailing platforms, affecting service reliability, driver utilization, customer experience, and platform efficiency. The objective of this study was to develop multiple machine learning models for predicting trip cancellation outcomes in ride-hailing platforms. The study adopted the Obtain, Scrub, Explore, Model, and iNterpret (OSEMN) framework and used 103,024 historical Ola and Uber booking records. Data preparation involved cleaning, handling missing values, feature engineering, encoding, and feature selection. Nine predictor variables were retained for model development: Vehicle Turn Around Time, Customer Turn Around Time, Customer Cancellation Reason, Driver Cancellation Reason, Ride Status, Payment Method, Ride Distance, Driver Ratings, and Customer Rating. The dataset was partitioned using an 80:20 stratified split, resulting in 82,419 training and 20,605 testing records. Class weighting was applied during model training to reduce bias towards the majority class. Four supervised machine learning models; Logistic Regression, Decision Tree, Gradient Boosting, and Extreme Gradient Boosting (XGBoost), were developed and evaluated using accuracy, precision, recall, F1-score, and confusion matrix. The models achieved training accuracies of 85.20%, 87.72%, 90.03%, and 92.52%, respectively. On the testing dataset, the corresponding accuracies were 84.00%, 86.02%, 89.06%, and 91.62%. XGBoost achieved the strongest testing performance, with 91.62% accuracy, 87.12% precision, 91.41% recall, and 89.12% F1-score. The relatively small differences between training and testing performance indicate that the developed models generalized well to unseen data. The study concludes that supervised machine learning can effectively predict trip cancellation outcomes, with XGBoost demonstrating the strongest predictive capability among the four developed models.
Institution: Technical University of Mombasa
Language: English
Subfield: Building and Construction
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7208816385