Cell Outage Detection and Degradation Classification Based on Alarms and KPI’s Correlation

Abstract

In this paper, we present cell outage detection and cell degradation classification algorithms for Self-Organizing Networks (SONs). The cell outage detection algorithm uses both the cell’s reported alarms and Key Performance Indicators (KPIs) to determine whether or not the cell is experience outage. For those cells that are not in outage, the cell degradation classification algorithm identifies the level of performance as either critical degradation, medium degradation or normal cell operation. The key idea of the proposed classification approach is to use the least number of KPIs by studying the correlation between the different KPIs. We consider three different machine learning algorithms for classification. Our results show that the Random Forest approach results in the highest accuracy of 99% with a runtime reduced by 29% due to the reduction in the number of used KPIs.