An efficient feature selection method for network video traffic classification
Yuning Dong, Quantao Yue, Feng Mao · 2017
A feature selection method RFPSO based on RelieiF and Particle Swarm Optimization (PSO) is proposed to mitigate the problem that the feature dimension of network traffic classification is too high. In this method, the ReliefF algorithm is used to filter out some irrelevant features and achieve the goal of rapid dimension reduction. Then, PSO is used as the search algorithm, and some better features are used as the partial initial population of particle swarm. The inconsistency rate is used as the evaluation function to select the optimal subset in the remaining feature subsets. The experimental results show that the classification accuracy of the RFPSO algorithm is higher than that of existing algorithms, and the computational complexity of the algorithm is lower than that of other two feature selection algorithms based on classification learning algorithm.