The Design of GBDT-based Anomaly Detection for O-RAN Near-RT RIC
Chih‐Cheng Tseng, Hsin-Cheng Wu, Shao‐Yu Lien, Jiun-Chi You, Changyi Liu, Rui-En You · 2025
With the increasing demand for the 5th Generation (5G) Base Stations (BSs), the global telecommunications industry is promoting technologies with open interfaces and open software and hardware designs, driving the 5G BSs to be cost-effectively developed. While such Open Radio Access Network (O-RAN) based 5G BSs are widely installed, User Equipment (UE) handovers become frequent and important. To instantly and accurately detect anomalous UE, a UE whose performance is below threshold, that needs to be handed over, Anomaly Detection (AD) becomes important. This paper applies the Gradient Boosting Decision Tree (GBDT) algorithm to detect anomalous UE. Based on the O-RAN architecture, the developed GBDT is implemented as an AD xApp. Specifically, using the RSRP and the available number of Physical Resource Blocks (PRBs) extracted from the performance metrics reported from the developed RAN emulator, this paper compares the performance of the developed GBDT-based AD xApp with the Isoforest-based AD xApp released by O-RAN Software Community (SC). Simulation results show the developed GBDT-based AD xApp demonstrates superior AD performance in terms of prediction accuracy and average prediction time. Specifically, when the number of UEs is 40, the prediction accuracy is improved by 102.2% and the average prediction time is reduced by 75.67%.