基于改进磷虾群优化的中心极大化KFCM算法在IDS的应用*

由于核模糊C-均值算法(Kernel Fuzzy C-means,KFCM)随机选择初始聚类中心,易导致算法陷入局部最优,且算法在聚类中心较近或重合时,易产生一致性聚类结果。为解决以上问题,提出一种改进算法。改进算法对原目标函数进行重新定义,通过在目标函数中增加一项聚类中心约束项来调控簇间分离度,从而避免算法出现一致性聚类结果。利用改进磷虾群算法对基于新目标函数的KFCM算法进行优化,使算法不再依赖初始聚类中心,提高算法的稳定性。基于距离最大最小原则产生多组较优的聚类中心作为初始磷虾群体并在算法迭代过程中融合一种新的精英保留策略,从而确保算法收敛到全局极值。通过对个体随机扩散活动进行分段式logistic混沌搜索,提高算法全局寻优能力。使用KDD CUP 99入侵检测数据进行仿真实验表明,改进算法具有更好的检测性能,解决了传统的聚类算法在入侵检测中稳定性差,检测准确率低的问题。

The Application Of Center Maximization KFCM algorithm Based On Improved Krill Herd in Intrusion Detection System

Due to the random selection of initial cluster centers,KFCM clustering algorithm is prone to local optimal. And when the clusteringScenters closeSor overlapp,Sthe algorithm is easy to produceSconsistent clusteringSresults.In order to solve the defects in kernel fuzzy C-means, an improved algorithm was proposed. The objective function is redefined by adding a cluster center constraint term to regulate inter-cluster separation, thus coincident clustering results are avoided. By using improved Krill Herd algorithm to optimize KFCM algorithm based on the new objective function, The problem of KFCM depending on initial center is effectively solved and the instability of clustering results is enhanced. The max-min distance method is used to produce manySexcellentScluster-