Enhancing QoE Prediction in 5G Video Streaming using Ensemble Learning Models

Ensuring smooth,reliable user experiences is essential, Whether we’re streaming videos, joining a Zoom call, playing games online, or browsing websites, we expect a smooth, enjoyable experience. That’s exactly where this Quality of Experience (QoE) Prediction Project steps in. With this rapid growth of multimedia applications and increasing demand for high-quality streaming services, accurately predicting Quality of Experience (QoE) has become a critical challenge in network and service management. Instead of waiting for the users to complain regarding buffering of videos or poor call quality, this system predicts their level of satisfaction that is known as MOS (Mean Opinion Score), before they even report a problem. That means service providers can act faster, fix issues, and deliver consistently high-quality service. This project proposes an intelligent machine learning based approach to predict QoE by analyzing various network, playback, and system-level attributes such as latency, throughput, jitter, packet loss, buffering time, and video resolution. The dataset collected contains detailed performance metrics recorded during video streaming sessions which are preprocessed and used to train multiple models. A Random Forest Regressor was used as the primary model to predict the MOS score, which is then categorized into QoE labels — Good, Average, or Poor, to make interpretation more user-friendly. This algorithm is great at handling complex data and give reliable predictions even when the data is noisy. Alternative models including K- Nearest Neighbors (KNN) which is simple yet effective model that mainly focuses on what similar users experienced to make decisions and the next algorithm is Support Vector Machine (SVM) which is excellent for drawing clear boundaries in data, especially when the relationship is not obvious, were also trained and evaluated for comparison. The final solution is integrated into a Flask-based web interface, allowing real-time QoE predictions based on user input.This system serves as a valuable tool for Internet Service Providers, content delivery platforms, and network engineers to proactively manage network resources and enhance user satisfaction.