A Supervised Learning Method for Anomaly Classification in Cable Networks Based on SNR Patterns

With the continuous expansion of communication networks, the number of electronic devices within these networks has grown substantially. As the number of devices and communication channels continues to rise, signal quality faces increasing threats, resulting in numerous network interruptions. This research paper is focused on cable network anomalies and the application of Machine Learning (ML) techniques to facilitate the maintenance of these networks. This study is based on real-world data collected from one of the largest Internet Service Providers (ISPs) that operates in Kosovo. This dataset is based on SNR (signal to noise ratio) values monitored during 100 different cases of outage incidents within the cable network. The SNR values are analyzed for each case, and factors that lead to such signal degradation are identified. The dataset also includes the necessary maintenance tasks that were required to address the outages. Using this dataset we apply a supervised learning technique to classify anomaly type by analyzing the SNR degradation patterns. The accuracy and reliability of the proposed method is tested using a separated part of the dataset.