Machine learning algorithms with Multi-Criteria-Decision-Making for cell outage detection in 5G / B5G

Self-organizing communication networks in 5G&B5G telecommunications are automatically optimized, configured and improved without human intervention, reducing costs (OPEX and CAPEX) in new generation networks. Self-healing is critical to 5G and B5G networks because it anticipates and resolves anomalies, optimizes performance, and adapts to changes in cell density and architecture. This process is done by managing out-of-service cells called COM. And COM itself is divided into two parts: COD to detect out-of-service cells and COC to compensate for out-of-service cells. Timely prediction of the work status of the cells increases the network performance. To achieve this goal, three law-based, algorithm-based and machine learning approaches are used in self-healing networks. This research uses combined machine learning algorithm based on multi-criteria decision making techniques to identify and estimate defective cells. Data were classified using a combination of three Bayesian learning algorithms, logistic regression method and AdaBoost classification method. The challenge of unbalanced training data was successfully solved with a hybrid approach and multi-criteria decision-based algorithm. Machine learning and TOPSIS improve efficiency and accuracy in decision-making and detection of defective cells.