A Supervised Machine Learning Mechanism for Traffic and Flow Control in LTE-A Scheduling

The provisioning of Quality of Service (QoS) for real-time applications in wireless networks is indeed a tricky task. Efficient network management poses some constraints given that the system has limited resources while traffic is competing for access. Besides, each application has different demands, and they need to be served according to their expectation. In this paper, we present a supervised machine learning mechanism based on the k-Nearest Neighbor (k-NN) algorithm to classify and select traffic in a general downlink scheduling procedure. The mechanism also implements a cross-layer communication approach between the MAC and application layer to control the transmission of recently served applications, reducing the overall load and alleviating the allocation process. This proposal can be added as a new stage in the most scheduling algorithms available in the literature. The mechanism has three basic steps: collection and labeling of the traffic data; classification; and flow control. The results obtained from simulation present good performance obtained for real-time applications considering measures of fairness, delay and packet loss.