IPTV Device Anomaly Detection: ML Approaches

There are many factors that affect customer satisfaction in IPTV services such as image quality, uninterrupted broadcasting, and loading time. The quality of customer service, technical support, and the problem-solving process also affect users’ trust in the service. In this study, it is aimed to increase customer satisfaction by minimizing the screen freezing error. Within the scope of the study, IPTV user data was analyzed with ML methods and device anomaly detection was performed. Screen freezing error was taken as the basis for device anomaly. In this direction, the STB, HGW, modem, OLT, and DSLAM devices in the IPTV structure where the error originated were analyzed with ML methods. The successes of the Isolation Forest, One Class SVM, and Local Disparity Factor models for device anomaly detection were compared with performance metrics. The Isolation Forest algorithm showed the best performance with an accuracy rate of 84%. With the analyzes performed within the scope of the study, it was determined which device model and which device card caused the screen freezing error, and it was ensured that device freezing errors were detected faster in operational processes.