基于模糊K-Means聚类的光纤大数据分类平台设计

Design of fiber-optic big data classification platform based on fuzzy K-Means clustering

光纤大数据分类容易出现稳定性差以及容易陷入局部最佳解问题,为了提高光纤大数据分类性能,设计基于模糊K-Means聚类的光纤大数据分类平台,平台数据采集器通过太网管制芯片CP2200和单片机C8051F340采集光纤大数据,并将数据传递给采用SOA体系结构的服务端为大数据分类提供服务基础,服务端通过传输接口将光纤数据传递给逻辑处理端,逻辑处理端接收到光纤数据后,通过云计算获取网络光纤大数据特征类型,同时结合软件设计的基于Witten框架的改进模糊聚类算法流程,制定光纤大数据分类标准,实现待分类光纤大数据的有效分类。平台软件给出基于Witten框架的改进模糊聚类算法分类光纤大数据的详细过程,先对光纤数据属性实施特征选择同时实施稀疏聚类,再融入空间距离代替传统欧氏距离,固定权值向量,不断优化聚类中心,获取准确的光纤大数据分类结果。实验结果说明,采用该平台分类高维光纤大数据的平均F-Measure值和平均准确率值分别达到0.9733和0.9641,分类性能佳。

Fiber optic data classification is prone to poor stability and easy to fall into local optimal solution.In order to improve the performance of fiber optic big data classification,a fiber optic big data classification platform based on fuzzy K-Means clustering is designed.The platform data collector collects fiber optic big data through the Ethernet control chip CP2200 and the single-chip microcomputer C8051F340,and passes the data to SOA architecture which provides the service foundation for big data classification.The server transmits optical fiber data to logical processing end through transmission interface.After receiving optical fiber data,the logical processing end obtains network optical fiber big data feature type through cloud computing.Based on improved fuzzy clustering algorithm flow of Witten framework,the optical fiber big data classification standard is formulated to realize the effective classification of optical fiber big data to be classified.The platform software gives a detailed process of classifying optical fiber big data based on the improved Witten framework fuzzy clustering algorithm.Firstly,the feature selection of optical fiber data is carried out and the sparse clustering is implemented at the same time,then the traditional Euclidean distance is replaced by spatial distance,fixing the weight vector,and optimize the cluster center continuously to obtain accurate optical fiber big data classification results.The experimental results show that the average F-Measure value and average accuracy rate of the high-dimensional optical fiber big data classified by this platform reach 0.9733 and 0.9641,respectively,and the classification performance is good.