基于LFKPCA-DWELM的入侵检测方案

基于机器学习的入侵检测系统普遍存在由于入侵数据维度大、数据样本不均衡和离散度大而严重影响分类性能的问题。提出了一种基于LFKPCA-DWELM的入侵检测算法，用改进的果蝇算法（LFOA）对核主成分分析算法（KPCA）进行优化，用优化后的核主成分分析算法（LFKPCA）对数据进行特征提取，将处理后的数据用于基于数据离散度的加权极限学习机（DWELM）的训练，最后使用训练好的模型进行分类实验。实验结果显示，该算法有效提高了检测率，降低了误报率和检测时间。

Intrusion Detection Algorithm based on LFKPCA-DWELM

The intrusion detection system based on machine learning generally has the problem that the classification performance is seriously affected by the large dimension of the intrusion data, the unbalanced data sample and the large dispersion degree. This paper proposes an intrusion detection algorithm based on LFKPCA-DWELM. First, the improved fruit fly algorithm（LFOA） is used to optimize the kernel principal component algorithm（KPCA）, and then the optimized kernel principal component algorithm（LFKPCA） is used to extract the features of the data. After that, it uses the processed data for training based on data dispersion beyond the extreme learning machine（DWELM）, and finally uses the trained model for classification experiments. Experimental results show that the algorithm can effectively improve the detection rate and reduce the false alarm rate and detection time.