Radio Transport Network (RTN) Alarm Detection Using Machine Learning: A Case Study of Ethio Telecom

Abstract This study proposes a machine learning-based framework to enhance the detection and prioritization of network alarms in Ethio Telecom's RTN device systems. It addresses challenges posed by the high volume of alarms generated by network components, which lead to inefficiencies and prolonged outages. 73,782 preprocessed alarms from the iMaster-NCE system of Ethio Telecom, covering three business regions (CNR, NER, and NEER), were utilized to develop alarm detection models. These models aim to detect business impact based on alarm severity, classify alarms effectively, and minimize system disruption. The framework integrates multiple machine learning algorithms, including k-nearest neighbor (KNN), decision tree (DT), and random forest (RF), within a stacking model architecture, with an LR meta-learner. Additionally, super vector machines (SVM) and Deep Learning (DL) algorithms are employed to compare the performance of stacking against SVM and DL using relevant evaluation metrics. This approach enhances model performance by leveraging the complementary strengths of individual classifiers. Evaluation metrics such as Accuracy, F1 score, Matthews Correlation Coefficient (MCC), and Area under the Curve (AUC) validate the model's robustness in handling imbalanced datasets and delivering reliable predictions. The results highlight the strong performance of the stacking model, SVM, and RNNs, displaying their effectiveness in classifying alarm severities with high precision and minimal misclassification. The proposed framework effectively filters critical alarms and reduces misclassification errors, enabling timely fault resolution. Additionally, integrating business area KPIs further improves predictive power and underscores the importance of domain-specific features in alarm management. The study highlights the potential of machine learning in optimizing network operations and suggests future research avenues, including the incorporation of additional KPIs, expanding to data from other vendors (e.g., Ericsson, ZTE), and utilizing deep learning techniques for handling complex datasets. This framework is practically applicable in telecommunications to ensure minimal system downtime and effective alarm detection.