Improved Multi-Granularity Cascade Forest Network Traffic Classification Optimization for SDN-Based Differentiated QoS Guarantees

How to achieve network quality of service guarantee and provide differentiated demand has been a hot research topic in the Internet field. Based on the data traffic classification problem of SDN (Software Defined Network) architecture, an improved multi-granularity cascade forest network traffic classification optimization model with differentiated QoS (Quality of Service) guarantee is proposed. First, the collected real dataset is normalized using the Z-score method, and the noise and edge samples are eliminated using the A-SMOTE algorithm. Then, the SDN-IgcForest model is proposed to improve the gcForest model by passing the low predictive confidence samples more accurately to the next level of cascaded forests for further training through the split-box method, and introducing the random sampling method to reduce the dimension and number of samples for generating the transformed features during the multi-granularity scanning. Finally, the experiments are conducted in the Mininet+Ryu simulation experiment environment. The results show that the pre-processing with A-SMOTE algorithm improves the classification performance by 8.09% over the SMOTE algorithm, and SDN-IgcForest performs better compared to DCNN, IgcForest and Proposed DFC.