QoS Prediction of 5G Network Slices using ML Models

Quality of Service (QoS) is a critical factor in 5G networks, ensuring reliable and efficient communication across diverse applications. In this paper, we introduce Machine Learning (ML) algorithms to predict QoS parameters in 5G network slices, enabling early detection and optimization of service degradation. Our system leverages advanced ML models to analyse key features like Slice Type, User Mobility, Slice Priority, Throughput, Jitter, and other network metrics without relying on manual monitoring. This application predicts QoS parameters and provides insights to improve network performance, ensuring better resource allocation and user experience. The system is trained to distinguish patterns from real-world and synthetic datasets, including parameters such as CPU utilization, memory usage, and environmental conditions (e.g., weather and time of day). Early detection of QoS degradation can prevent major service disruptions, including reduced network reliability, increased latency, and poor user experience in critical applications like autonomous vehicles and telemedicine. Traditional methods for monitoring and optimizing QoS often require manual intervention and substantial computational resources. These approaches may lack the responsiveness and scalability required for modern 5G networks. Our proposed system uses an efficient, automated, and adaptive method to predict QoS by analysing multiple factors simultaneously. The application provides network operators with real-time insights and actionable recommendations to maintain optimal service quality. The process is implemented using advanced ML models, including Random Forest, Linear Regression model and Decision Tree Model, with Random Forest providing the best performance through hyperparameter tuning. We utilize an enhanced QoS dataset collected from Kaggle and enriched with synthetic data to simulate real-world 5G conditions. After dataset preprocessing— such as feature extraction, scaling, and encoding—model training is conducted, and the final optimized model is deployed using Django for real-time predictions and analysis.