A comparative analysis of supervised learning techniques for level of service classification and delay prediction at signalized intersections
Engineering Applications of Artificial Intelligence, vol.181, 2026 (SCI-Expanded, Scopus)
- Publication Type: Article / Article
- Volume: 181
- Publication Date: 2026
- Doi Number: 10.1016/j.engappai.2026.115504
- Journal Name: Engineering Applications of Artificial Intelligence
- Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, INSPEC, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO)
- Keywords: Artificial intelligence, Delay analysis, Level of service, Machine learning, Signalized intersections, Supervised learning
- Dokuz Eylül University Affiliated: Yes
Abstract
Signalized intersections play a critical role in urban traffic efficiency, where delay and level of service (LOS) are key performance indicators for improving operational efficiency. This study presents a data-driven framework based on artificial intelligence and machine learning techniques to predict delay and classify level of service using traffic and geometric features at three-leg signalized intersections. The proposed approach integrates regression and classification models, including extreme gradient boosting (XGBoost), random forest, and light gradient boosting machine (LightGBM), while interpretability is enhanced through feature importance analysis, Shapley value-based attribution, and model distillation techniques. The results show that tree-based ensemble methods, particularly XGBoost, achieve strong predictive performance by capturing nonlinear relationships in both delay estimation and LOS classification, while LightGBM exhibits comparable performance in classification tasks. To enhance interpretability, the LightGBM model is distilled into a classification and regression tree (CART) model that preserves key decision patterns. This framework provides an interpretable decision-support system to achieve effective signal control and planning strategies and to evaluate operational performance under varying traffic and geometric conditions via supervised learning approaches.