构建基于密度峰值聚类算法的反作弊系统  

Anti-fraud System Based on Density Peak Clustering Algorithm

作　　者：叶楠 寇丽杰 YE Nan;KOU Lijie(Fuzhou University of Technology,Fuzhou 350506,China)

机构地区：[1]福州理工学院,福建福州350506

出　　处：《数字通信世界》2022年第5期46-48,90,共4页Digital Communication World

基　　金：2019年度中青年教师教育科研项目(JAT191014)。

摘　　要：文章针对密度峰值聚类算法(Densit y Pe ak,DPe ak)能够有效解决离群点检测和分类的问题,结合算法特点进行改良优化,提出了一种基于密度峰值聚类算法的反作弊模型,通过构建对抗性学习框架、优化算法计算复杂度、引入LOF判别分类机制、多模态融合技术等创新方法,实现了不同阶段黑产演变环境下的规则校验和分类判定。该模型有效解决了百万级数据集时间复杂度高、任意形状类簇的聚类等流量反作弊服务的领域性挑战问题,对于密度峰值聚类算法的产业应用具有一定的研究意义。实验结果表明,反作弊模型在聚类效果和算力测试上均达到了预期结果,证明了该模型的有效性。For the density peak clustering algorithm(density peak,DPeak)can effectively solve the problem of detection and classification of outliers,combining with the characteristics of algorithm improved optimization,this paper proposes a clustering algorithm based on density peak anti-cheat model,by building a confrontational learning framework,optimization algorithm computational complexity,introducing LOF discriminant classification mechanism,mult-imodal fusion technology innovation methods,realizes the different stages of black production environment and evolvement rule check and classification.This model can effectively solve the domain challenges of high time complexity of millions of data sets,clustering of arbitrary shape clusters and so on.It has a certain research significance for the industrial application of peak density clustering algorithm.The experimental results show that the anti cheating model achieves the expected results in clustering effect and computational power test,which proves the effectiveness of the model.

关 键 词：密度峰值 聚类算法 截断距离 反作弊 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]