Prediction of Round-Trip Time in 5G NSA Networks in Urban Environment

Latency is a critical component in influencing application performance and user experience, and it is widely measured through Round Trip Time (RTT). In this work, we aim at predicting application-layer latency from radio- and physical-layer information performing binary classification with Machine Learning (ML) algorithms. Leveraging radio layer parameters, we perform binary classification to verify whether RTT is above or below a chosen threshold following a specific methodology to guarantee user requirements for real time application to be met. We discovered that the best performing ones are the ensemble methods such as Decision Tree, Random Forest and Gradient Boosting Classifiers. The radio-layer parameters which appear to be more relevant are power and quality radio layer parameters as Synchronization Signal Reference Signal Received Power (SS-RSRP), Synchronization Signal Reference Signal Received Quality (SS-RSRQ) and Synchronization Signal Signal-to-Interference-plus-Noise Ratio (SS-SINR) as shown by P-value and F-Score.