基于机器学习的光纤故障大数据分类算法研究  

Research on Classification algorithm of Optical Fiber Fault Big Data Based on Machine Learning

作　　者：张华[1] ZHANG Hua(Department of Computer and Information Security Management,Fujian Police College,Fuzhou 350007,China)

机构地区：[1]福建警察学院计算机与信息安全管理系

出　　处：《安阳工学院学报》2019年第6期52-55,107,共5页Journal of Anyang Institute of Technology

摘　　要：为了提高光纤故障诊断能力,提出一种基于机器学习的光纤故障大数据分类算法。对采集的光纤网络传输数据进行低维度的特征集构造,构建光纤故障分布大数据库,对光纤故障数据库中的异常数据特征集采用加权统计分析方法进行样本回归分析;以少量的样本类别数据为测试集,采用层次聚类方法对光纤故障大数据进行极端随机数分析;对全部的故障样本进行抽样训练,提取光纤故障大数据的关联特征量,结合K-means算法和最近邻算法进行数据聚类中心扰动性分析;将光纤故障大数据的特征提取结果输入到机器学习器中进行数据分类,结合大数据融合聚类方法实现光纤故障大数据分类。仿真结果表明,采用该方法进行光纤故障大数据分类的准确性较高,误分率较小,提高了故障检测识别能力。In order to improve the capability of optical fiber fault diagnosis, a classification algorithm of optical fiber fault big data based on machine learning is proposed. To construct the low-dimensional feature set of the collected optical fiber network transmission data, to construct the large database of optical fiber fault distribution,and to use the weighted statistical analysis method to carry out sample regression analysis to the abnormal data feature set in the optical fiber fault database. Taking a few sample category data as test set, using hierarchical clustering method to carry on extreme random number analysis to big data of optical fiber fault, using all fault samples to carry on sampling training, extracting the correlation characteristic quantity of optical fiber fault big data. Combining the K-means algorithm and the nearest neighbor algorithm to analyze the central disturbance of data clustering, the feature extraction results of big data are input into the machine learning device for data classification. In this paper, big data fusion clustering method is used to realize the classification of optical fiber faults big data. The simulation results show that this method has high accuracy and low misclassification rate in the classification of optical fiber fault big data, and improves the ability of fault detection and identification.

关 键 词：机器学习 光纤故障大数据 分类算法 特征提取 

分 类 号：TN929.1[电子电信—通信与信息系统]