Anomaly detection of the tracking performance of a Toptracer range installation : Exploration of various methods

Abstract: This thesis work examines anomaly detection methods on data sets related to sports, more especially related to the game of golf. This thesis aims to detect sensor malfunction on a TopTracer range installation. Different methods and modeling frameworks are examined. Supervised and unsupervised machine learning methods and statistical methods are explored, and other paths are mentioned briefly. Unfortunately, no positive results on anomaly detection could be drawn using clustering or classification algorithms, but a final method using a handmade anomaly score showed some positive results. The final method is not satisfactory but gives a clear idea of future works to improve it, and the different paths explored considerably reduced the panel of anomaly detection techniques related to the thesis-specific problem.