A Scoring Framework for the Unsupervised Identification of the Relevance of Metrics in Cell Degradation Clusters

The complexity of cellular networks makes them difficult to manage and prone to failures. This has led to the application of artificial intelligence mechanisms for failure management tasks. Nevertheless, the availability of labeled data on failure cases is limited, making unsupervised techniques the most relevant to apply. However, these techniques require network experts to analyze the results through a costly process to link them to specific cases. Intending to ease the labeling of cases, the present work proposes a framework to determine the relevant metrics for any clustering. Furthermore, the framework is evaluated using data from real cellular networks.