Active Build-Model Random Forest Method for Network Traffic Classification

Network traffic classification continues to be an interesting subject among numerous networking communities. This method introduces multi-beneficial solutions in different avenues, such as network security, network management, anomaly detection, and quality-of-service. In this paper, we propose a supervised machine learning method that efficiently classifies different types of applications using the Active Build-Model Random Forest (ABRF) method. This method constructs a new build model for the original Random Forest (RF) method to decrease processing time. This build model includes only the active trees (i.e., trees with high accuracy), whereas the passive trees are excluded from the forest. The passive trees were excluded without any negative effect on classification accuracy. Results show that the ABRF method decreases the processing time by up to 37.5% compared with the original RF method. Our model has an overall accuracy of 98.66% based on the benchmark dataset considered in this paper. Keyword-Network Traffic Classification, Machine learning, Supervised Learning, Random Forests Algorithm