基于K均值和k近邻的半监督流量分类算法

Semi-Supervised Traffic Identification Based on K-Means and k-Nearest Neighbours

针对流量分类中样本标注瓶颈和类不均衡问题,提出一种基于K均值和k近邻的半监督流量分类算法。采用K均值聚类算法将混有少量标记样本和大量未标记样本的数据聚成若干个簇,然后采用k近邻算法利用簇中标记样本对未标记样本进行分类。在分类过程中根据簇中标记样本分布调整参与分类的最近邻居数,从而克服了类不均衡对识别小类流的不利影响。理论分析和实验结果都表明,算法在面对非均衡协议流时提高了小类流的识别率。

In order to solve the problem of sample marking bottleneck and the imbalanced protocol flow,a semi-supervised traffic identification algorithm based on K-means and k-Nearest Neighbour （kNN）＇ is presented. The K-means algorithm is first employed to partition a training dataset that consists of a few labeled flows combined with abundant unlabeled flows. Then, the unlabeled smaples are identified using the labeled samples in the cluster based on kNN. The number of the nearest neighbours is adjusted according to the distribution of the labeled samples in the cluster,which over- comes the adverse effects of the imbalanced protocol flows to identify the minority flows. Theoretical analysis and experimental results show that the algorithm can improve the recognition rate of minority flows in the ease of the imbalanced protocol flows.