自适应权重优化的树突状细胞算法  

Adaptive Weight Optimization of Dendritic Cell Algorithm

作　　者：田雨柔 杨鹤[2,3] TIAN Yu-rou;YANG He(College of Computer and Information Engineering,Hubei Normal University,Huangshi Hubei 435002,China;School of Computer Science,Hubei University of Education,Wuhan 430205,China;Hubei Collaborative Innovation Center for Basic Education Information Technology Service,Wuhan 430205,China)

机构地区：[1]湖北师范大学计算机与信息工程学院,湖北黄石435002 [2]湖北第二师范学院计算机学院,武汉430205 [3]基础教育信息技术服务湖北省协同创新中心,武汉430205

出　　处：《湖北第二师范学院学报》2024年第8期43-51,共9页Journal of Hubei University of Education

摘　　要：树突状细胞算法是人工免疫系统中先天免疫层的经典算法,该算法通过融合危险和安全信号发现异常。但树突状细胞算法在信号权重的取值上常需要根据数据特征手动设置,削弱了算法的自适应性。针对这一问题,引入了网格搜索法,在给定的权重范围内根据识别效果自动调整权重取值,得到适应不同类型和规模数据集的信号权重组合。在多个公开数据集上的实验结果表明自适应权重优化的树突状细胞算法能根据数据集特征自适应训练得到较合理的权重矩阵,减少了人工经验对算法准确率的影响,改进后的算法识别准确率、真阳性率等均高于原树突状细胞算法,并优于同类算法。Dendritic Cell Algorithm is a classic algorithm in the innate immune layer of artificial immune systems,which detect anomalies by integrating danger and safe signals.However,the dendritic cell algorithm often requires manual setting of signal weights based on data features,thereby diminishing its adaptability.To address this issue,a grid search method which can automatically adjust weights within a given range based on detection accuracy has been introduced in order to obtain adaptive signal weights suitable for different types and scales of datasets.Experimental results on several public datasets demonstrated that the adaptive weight optimized dendritic cell algorithm can train reasonable weight matrices according to characteristics of dataset,thus reducing the impact of human experience on algorithm accuracy.The improved algorithm exhibits higher detection accuracy and true positive rates than the original dendritic cell algorithm and perform better than similar algorithms.

关 键 词：自适应权重 网格搜索 树突状细胞算法(DCA) 人工免疫系统 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]