改进粒子群联合禁忌搜索的特征选择算法

Feature selection algorithm based on improved particle swarm joint taboo search

针对入侵检测中数据特征维度高的问题,提出了改进粒子群联合禁忌搜索(IPSO-TS)的特征选择算法。采用遗传算子对粒子群算法进行了改进,得到了特征选择初始最优解;对该解进行禁忌搜索(TS)得到了特征子集的全局优化解。基于KDD CUP 99数据集的实验结果表明,相较遗传算子整合粒子群算法(CMPSO)、粒子群算法(PSO)和粒子群联合禁忌算法,IPSO-TS减少了至少29.2%的特征,缩短了至少15%的平均检测时间,提高了至少2.96%的平均分类准确率。

To solve the problem of high data feature dimensionality in intrusion detection,a feature selection algorithm based on improved particle swarm optimization taboo search(IPSO-TS)was proposed.The genetic algorithm was used to improve the particle swarm optimization,and the initial optimal solution of feature selection was obtained.A taboo search(TS)algorithm was used for initial optimal solution to obtain the global optimal solution of the feature subset.Compared with genetic algorithm integrated particle swarm optimization(CMPSO),particle swarm optimization(PSO)and PSO-TS algorithms,experimental results based on the KDD CUP 99 dataset show that the method reduces the features by about 29.2%,shortens about 15%of the average detection time,and increases about 2.96% of the average classification accuracy.