基于SOM聚类的物联网大数据中有效信息挖掘系统  

Effective information mining system in IoT big data based on SOM clustering

作　　者：邓凯 章荣燕 郭清 李宇 陈隆晖 徐靖淞 DENG Kai;ZHANG Rongyan;GUO Qing;LI Yu;CHEN Longhui;XU Jingsong(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6期53-56,62,共5页Electronic Design Engineering

摘　　要：针对物联网大数据中有效信息挖掘困难的问题,对其根源进行分析,该问题主要是数据资源分配不清晰导致的。因此提出结合粒子群算法对SOM聚类进行改进的物联网大数据有效信息挖掘系统。通过粒子群算法对SOM聚类的权值进行优化,并结合自回归模型对数据特征作出估计,同时对集群进行动态分配。经过实验验证,结果表明改进后的算法的资源利用率更高,对数据特征的预测更加准确,有效信息的挖掘效率更高,整体上的执行延迟在0.2 ms左右。Aiming at the problem of difficulty with effective information mining in big data of the Internet of Things,the root cause is analyzed,which is mainly caused by the lack of clarity in the allocation of data resources.Therefore,an effective information mining system for Internet of Things big data combined with particle swarm algorithm to improve SOM clustering is proposed.The weight of SOM clustering is optimized by particle swarm algorithm,and the data features are estimated by combining with the autoregressive model,and the clusters are dynamically assigned.The experimental results show that the improved algorithm has higher resource utilization,more accurate prediction of data features,higher mining efficiency of effective information,and an overall execution delay of about 0.2 ms.

关 键 词：物联网 数据流 粒子群算法 SOM聚类 数据特征 数据信息 

分 类 号：TN98[电子电信—信息与通信工程]