基于动态网格的非平衡大数据密度聚类方法  

Unbalanced big data density clustering method based on dynamic grid

作　　者：郭清 李睿 李宇 章荣燕 刘伟 雷宇 GUO Qing;LI Rui;LI Yu;ZHANG Rongyan;LIU Wei;LEI Yu(Guizhou Tobacco Redrying Co.,Ltd.,Guiyang 550005,China)

机构地区：[1]贵州烟叶复烤有限责任公司,贵州贵阳550005

出　　处：《电子设计工程》2025年第3期162-167,共6页Electronic Design Engineering

摘　　要：针对非平衡大数据当中进行聚类较为繁琐且聚类结果准确度不高的问题,提出一种以动态网格为基础的密度聚类方式。通过动态网格的划分,并设置相应网格密度的阈值,进行网格的自适应生成,实现相应的密度聚类效果。算法通过样本训练与测试对用户的异常轨迹进行监测,提出类相似的概念对不同的格簇进行划分,同时将噪声当成异常数据进行检测,保证数据检测的全面性。经过实际实验验证,改进算法对于非平衡大数据等问题的处理效果更优,精确度更高。Aiming at the problem of cumbersome clustering and low accuracy of clustering results in imbalanced big data,a density clustering method based on dynamic grids is proposed.By dividing the dynamic grid and setting a threshold for the corresponding grid density,adaptive grid generation is carried out to achieve the corresponding density clustering effect.The algorithm monitors the abnormal trajectories of users through sample training and testing,proposes the concept of class similarity to partition different lattice clusters,and detects noise as abnormal data to ensure the comprehensiveness of data detection.After actual experimental verification,the improved algorithm has better processing effect and higher accuracy for problems such as imbalanced big data.

关 键 词：动态网格 非平衡大数据 数据流 类相似 异常轨迹 

分 类 号：TN-9[电子电信]