一种基于并行免疫网络的大数据分类算法

A classification algorithm for big data based on parallel immune network

为了解决数据量增加时串行免疫网络算法难以实现大数据处理的问题,提出了并行免疫网络训练和分类模型,并在Spark并行框架下设计了并行免疫网络分类算法.给出了入侵检测大数据背景知识;建立了Ainet并行算法框架,详述了并行免疫网络分类算法步骤;采用cup99入侵检测数据集进行了试验,进而将并行Ainet算法同其他算法做了比较.试验结果表明:较串行Ainet算法,并行Ainet算法训练时间下降了11/12,检测时间降低了19/20,准确率提高了10%,同时检测率提高了5%,而误报率降低了20%,可见并行Ainet算法各方面都取得较好的效果;试验验证了分类效果对训练数据集数量敏感的特点;并行Ainet算法在准确率、检测率和误报率方面优于其他算法,但运行时间较长.

To solve the problem of processing difficulty for big data in the serial immune network algorithm,the parallel immune network training and classification model were proposed,and a parallel immune network classification algorithm was designed under the framework of parallel Spark.The background knowledge of intrusion detection big data was introduced to establish Ainet parallel algorithm framework,and the algorithm steps of the proposed algorithm were described in detail.The cup99 intrusion detection data set was adopted in the experiments,and the Ainet algorithm was compared with other algorithms.The experimental results show that compared with the serial Ainet algorithm,the parallel Ainet algorithm can reduce the training time by 11/12 and the detection time by 19/20 and can improve the accuracy by 10%and the detection rate by 5%with reduced false alarm rate of 20%.The parallel Ainet algorithm achieves good effect in all aspects.The experimental verification of classification illuminates that the number of training data set has sensitive feature.The parallel Ainet algorithm outperforms other algorithms in accuracy,detection rate and false alarm rate with poor run time.