Probabilistic Regression and Anomaly Detection for Latency Assessment in Mobile Radio Networks

This thesis provides a thorough examination and empirical results on the use of machine learning for predicting latency in mobile radio networks, specifically emphasizing probabilistic regression and anomaly detection tasks. After a ML-aided selection of the Key Performance Indicators that most influence the latency, different models are compared for both probabilistic regression and anomaly detection. Such models present network designers with a valuable instrument to explore the correlations that exist between particular network Key Performance Indicators and latency.