Anomaly Detection of Internet Service Quality Degradation in Digital Twin for Fixed Access Network

The Internet service quality of fixed broadband has changed from the access rate to the concern of user perception of Internet access, but it is difficult to collect the service perception indicator. At the same time, the fixed broadband services are long process services, involving many network elements, indicators. The data of these indicators are scattered in different systems, so there is a certain difficulty in data collection, data standardization, quality assurance, also including the abnormal diagnosis of fixed broadband services. In this paper, for anomaly detection of Internet service quality degradation, we propose to establish a digital twin system for quasi-real-time simulation of full indicators analysis and intelligent operation. The twin system supports indicators aggregation analysis, data feature extraction, anomaly detection and network fault diagnosis. Through the quasi-real-time synchronous online training model, it evaluates the state of the multi-level network of the fixed broadband, finds network abnormalities, and provides timely feedback to operation and maintenance personnel to ensure reliable operation of the fixed broadband. The intelligent operation model in digital twin system is verified using real data from the existing network, and the accuracy of abnormality detection achieves the industry's leading level. The system will subsequently be docked with the actual network operation and maintenance system, thus realizing real-time abnormality discovery of the system.