Packet Classification Using Improved Random Forest Algorithm

Packet classification in Software-Defined Network (SDN) is a bottleneck process that falls in the critical path of the packet processing in network system. To increase the performance of packet processing, the processing time of the forwarding engine should be enhanced. As the height of the tree is increased, the depth of the hidden layer is increased which leads to the problem of higher time complexity. In this article, an efficient time constraints machine learning technique, namely Modified Random Forest (MRF) algorithm, is proposed. This algorithm increases the performance accuracy and the time required to perform classification is minimum when compared to DNN classifier. The modified random forest classifier classifies the packet more accurately in less computational time.