Explainable AI and Statistical Learning for Enhanced Abnormal Detection in O-RAN Networks

This paper focuses on the critical task of abnormal detection in the Open Radio Access Network (O-RAN) environment. Utilizing real-world O-RAN data, we propose a novel abnormal detection method that integrates explainable Artificial Intelligence (AI) with statistical learning techniques. Our approach classifies O-RAN devices based on their operational locations and utilizes historical data for AI training, emphasizing the identification of deviations in network performance metrics such as Radio Resource Control (RRC) connections. Through field testing, the feasibility of our method has been validated. This study provides a scalable and transparent abnormal detection solution for network management in O-RAN systems, highlighting the importance of applying explainable AI in the field of telecommunications.