Classification-based approach for cell outage detection in self-healing heterogeneous networks

Future mobile wireless communication networks will be featured as heterogeneity in order to enhance network performance and improve user experience. For better adaption to network challenges over its complexity and vulnerability, cell outage detection technique, a promising intelligent part of self-organizing networks (SON), has drawn considerable attention to deal with unexpected network faults. Our work is devoted to cell outage detection in a two-tier macro-pico network. Based on observation of performance metrics in time domain, we employ a classification algorithm called K-nearest neighbor (KNN) to achieve automatic anomaly detection. With some reasonable assumptions and a LTE-A system simulator, numerical experiments are implemented to demonstrate the efficiency of the proposed algorithm. Finally, localization for anomaly data and performance evaluation are further carried out to validate the classification accuracy.