A Novel Framework for QoS Prediction of V2X in 5G and B5G Networks: a Unified Approach with Explainable Artificial Intelligence (XAI) and Nested Cross-Validation

Prediction of Quality of Service (QoS) in Vehicle-to-Everything (V2X) communication is essential for ensuring reliable and efficient data exchange within connected and cooperative autonomous vehicles. By predicting QoS, it is possible to maintain the necessary communication standards, even in dynamic and unpredictable environments. With the advents in 5G and B5G technologies, the need for accurate QoS prediction models are expected to be crucial to ensure the reliability, safety, and effectiveness of V2X applications as well as other vertical applications. In this paper, we propose a novel and unified machine learning (ML) framework for predicting Packet Delivery Ratio (PDR) and Throughput, combining nested cross-validation scheme with eXplainable AI (XAI) techniques. Our methodology composes of training and testing three different ML models, and compare them in terms of Root Mean Square Error metric. Although training with nested cross-validation scheme requires longer computational times compared to other validation approaches, it achieves better performance with lower error rates. Furthermore, we integrate various XAI methods within the nested cross-validation scheme by combining the results from different models, enhancing the interpretability and reliability of our predictions.