Contextual classification of network traffic through rough set theory and convolutional neural networks

Now-a-days online activities and the usage of various applications by users are more diverse and led to the tremendous increase of network traffic. Monitoring and analyzing of the network became an issue due to the increased traffic. This paper deals with classifying network channels for configuring networks for optimum utilization and uninterrupted soft services to the users. The recognition of user context must be taken into consideration to improve the user experience on the network. This paper proposes a new feature selection method using Rough Set Theory (RST) and high-accuracy classification using Convolutional Neural Networks (CNN), which demonstrate a good performance. RST is responsible for the quality improvement of features fed to the CNN that results in the improvement of classification accuracy. Moreover, a new optimization technique called Grey Wolf sweep is added that tunes four CNN settings. The final model hits 0.962 macro-F1 and labels each packet in 0.22 milliseconds (ms). The evolution of this technology results in the development of responsive network infrastructures that can change their conditions according to user requirements. The combination of RST and CNN of our work as revealed a significant effort in the field of network management to apply machine learning for the traffic analysis in networks.