LAUSR

Self-organizing networks (SONs) aim at automating the management of cellular networks. However, tasks, such as the selection of the most appropriate performance indicators for SON functions, are still carried out by experts. In this letter, an unsupervised and autonomous technique for the selection of the most useful performance indicators is proposed, consisting in a data clustering stage followed by a supervised procedure for feature selection. Results show that the proposed method effectively relieves and outperforms an expert’s selection, allowing a drastic reduction of the volume and complexity of both network databases and SON procedures without human intervention.