光纤网络数据的聚类算法改进  

Improvement of clustering algorithm for fiber network data

作　　者：金红军[1] 王正平[2] JIN Hongjun;WANG Zhengping(Yancheng Teacher’s University,Yancheng Jiangsu 224002,China;Nanjing Normal University,Nanjing Jiangsu 210097,China)

机构地区：[1]盐城师范学院,江苏盐城224002 [2]南京师范大学,南京210097

出　　处：《激光杂志》2019年第12期180-183,共4页Laser Journal

基　　金：江苏省高校自然科学研究项目(No.14KJB520040);江苏省社会科学基金(15JYB009)

摘　　要：传统光纤网络数据具有容量大和分散性强的特点,导致挖掘的聚类性不好。提出基于联合参量估计的光纤网络数据的聚类算法。分析光纤网络数据的存储结构模型,结合数据结构的优化重组和自适应调度方法进行光纤数据的离散化特征重构,提取光纤数据的关联特征,采用联合参量估计法方法进行模糊信息聚类处理,根据数据聚类结果,结合先验样本迭代训练方法,实现对光纤网络数据的挖掘。仿真结果表明,采用该方法进行光纤存储数据挖掘的模糊聚类性较好,收敛距离缩短且时间开销分别缩短了0.33 s和9.74 s,聚类能力较强。Traditional optical fiber network data has the characteristics of large capacity and strong dispersion,resulting in poor clustering of mining.A clustering algorithm for optical fiber network data based on joint parameter estimation is proposed.The storage structure model of optical fiber network data is analyzed.The discretization characteristics of optical fiber data are reconstructed by combining the optimization and reorganization of data structure with the adaptive scheduling method.The correlation characteristics of optical fiber data are extracted.The fuzzy information clustering is processed by the joint parameter estimation method.According to the data clustering results,the data mining of optical fiber network is realized by combining with the prior sample iterative training method.The simulation results show that the method has better clustering ability,shorter convergence distance and shorter time overhead by 0.33 s and 9.74 s respectively.

关 键 词：光纤数据 聚类算法 模糊聚类 特征提取 

分 类 号：TN912[电子电信—通信与信息系统]