An accurate traffic classification model based on support vector machines

摘要: Network traffic classification is a fundamental research topic on high-performance network protocol design and operation management. Compared with other state-of-the-art studies done the classification, machine learning ML methods are more flexible intelligent, which can automatically search for describe useful structural patterns in supplied dataset. As typical method, support vector machines SVMs based statistical theory has high accuracy stability. However, performance of SVM classifier be severely affected by data scale, feature dimension, parameters classifier. In this paper, real-time accurate training model named SPP-SVM proposed. An deducted from scaling dataset employs principal component analysis PCA to extract features verify its relevant obtained PCA. By employing algorithm do dimension extraction, confirms critical features, reduces redundancy among them, lowers original so as reduce over fitting increase generalization effectively. The optimal working kernel function used derived improved particle swarm optimization algorithm, will optimize global solution make inertia weight coefficient adaptive without searching wide range, traversing all parameter points grid adjusting steps gradually. two- multi-class classifiers proved 2 sets traces, coming different topological Internet. Experiments show that SPP-SVM's superior supervised algorithms performs significantly better than traditional accuracy, elapsed time.