Abstract
Reliable public transport monitoring is an important component of smart city mobility services. Although Internet of Things (IoT)-based GPS monitoring provides extensive data on vehicle movements, technical interference and sensor errors often make it difficult to detect operational anomalies. This paper presents a two-stage method for cross-validating automated anomaly detection with passenger-generated complaints within an urban bus network, organizing the data into technical and operational domains before resolving them into granular categories. Using a dataset of 1.7 million GPS points and 116 manual reports across three routes, we analyzed four main types of anomaly: traffic jam, GPS anomaly, stop overstay, and bus skip. Our results demonstrate that integrating crowd-sourced data improves the agreement rate of detection models by 1.8%, reaching an agreement rate of 9.4% in identifying service-related issues. The study highlights the disparity between technical GPS artifacts and passenger-perceived service quality, providing a foundation for more resilient urban mobility management.