Diagnosis Based on Machine Learning for LTE Self-Healing

Self-Healing is one of the most important parts of self-organizing communication networks that offer detecting cells with a service of degradation, finding out the fault cause, and executing compensation and repairing actions. Diagnosis identifying the fault caused by problem cells is one of the most complex tasks. To perform the diagnosis, this paper presents two multi-classification diagnosis system based on machine learning methods, namely SVM (Support Vector Machine) and AdaBoost (Adaptive Boosting). Results show that the performance of the AdaBoost method compared to SVM with linear kernel is significantly better in terms of diagnosis error rate and undetected rate, but the false positive rate of AdaBoost is a little higher than SVM. SVM is more focusing on filtering normal cases, but AdaBoost is more inclined to find out fault cases. It indicates that the diagnosis system based on AdaBoost has high accuracy and reliability than SVM in this data set.