Traffic Classification for Network Slicing Using Machine Learning Techniques

With the advent of 5G networks, cities will become more connected and intelligent, and the Internet of Things (IoT) will become an increasingly present reality in society's daily life. 5G networks promise to support a wide range of services, which will result in an exponential increase in network traffic over the next few years. Given the scale and complexity expected for 5G networks, where networks are flexible and their services are instantiated on demand, there is no time to react to a situation of congestion and other changes in network behavior. Knowing IP traffic patterns will be an essential aspect for the effective management of 5G networks. By incorporating machine learning techniques into network management, it becomes possible to analyze demand trends and anticipate them, as well as making it possible to allocate a more suitable slice of the network to serve a particular group of applications. In this context, this work provides a detailed analysis of IP traffic, based on statistical data extracted from a real network, using Machine Learning (ML) techniques. This study contributes to identifying the attributes that most influence the classification of IP traffic. The results show that the Random Forest model trained with only 5 statistical features achieves excellent performance for the Precision, Recall and F1-score metrics and high accuracy in all classes. In addition, we propose an approach that uses ML models for the dynamic resource allocation for Network Slices.