Data-Driven and Evidence-Based Computing Algorithms to Detect Anomalies of Cycling Facilities

Abstract This paper explores the development of a data-driven, evidencebased approach to manage the quality of cycling facilities, focusing on detecting anomalies based on vibration data collected from different road types. We introduce a method for reconstructing normal vibration patterns from mixed datasets, demonstrating the model’s ability to capture smooth and bumpy road conditions, as well as the challenges posed by sloped surfaces. A statistical threshold for anomaly detection is defined by combining accelerometer, gyroscope, and speed data, providing a comprehensive approach to identifying anomalous vibrations. Results show that the computing model can effectively distinguish between normal and anomalous data across various road conditions, offering a robust solution for managing cycling infrastructure with real-time, data-driven insights.