Anomaly Detection and Root Cause Analysis Enabled by Artificial Intelligence

Traditional anomaly detection and root cause analysis in radio access network are not accurate and efficient enough to enable automatic network operation and maintenance in large-scale 4G/5G heterogeneous network. In this paper, we designed a framework of anomaly detection and root cause analysis, which was used in the field trial. In the framework, we proposed an algorithm of anomaly detection together with a labeled dataset of radio access network. The algorithm is proved to be more efficient with slightly performance loss compared with the existing State-of-the-Art algorithm. In addition, the automatic network closed-loop method for capacity problem, including the proposed anomaly detection and root cause analysis, has been verified in several base station sites based on O-RAN architecture.