Deep learning-based quality of experience (QoE) classification for optimal resource allocation in manufacturing networks

Abstract

With the ever-changing environment in industrial 4.0, ensuring a high Quality of Experience (QoE) for users would require network traffic data to be efficiently managed to maintain an optimal system performance. This project proposes a hybrid deep learning approach using a Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM) model to classify network traffic within a Multi-access Edge Computing (MEC) framework. By using the traffic data from the captured Packet Capture (PCAP) files, the features engineered were used for classification. The engineered features include packet level and flow level data such as throughput, jitter and payload characteristics. Extensive experimentation was conducted using feature selection techniques, hyperparameter tuning, and K-Fold cross-validation, obtaining a classification accuracy of 89.1%. The model was further evaluated for its adaptability to broader use cases such as Industrial Internet of Things (IoT) and cybersecurity, in order to highlight its potential for real-world deployment in such dynamic network environments.