摘要
数据流断点区检测用于挖掘识别动态数据流中的异常数据,由于移动网络数据流具有无限输入与动态变化等特征,针对现有算法存在对动态数据的聚类精度不高、离群点的处理效率较差的缺陷,难以实现移动网络环境下数据流断点区的实时高精度检测,提出了一种改进加权近邻密度的动态数据流断点区优化检测算法。上述算法利用微簇的密度来发现数据流中的离群点与簇连接,用以表示动态数据流的演化信息,为了更准确体现数据流的原始特征,在算法模型中引进了动态簇更新维护机制,并且通过改进权值邻近算法,当计算模型检测出有新对象类产生,根据聚类算法重建计算模型,这样既能准确实时发现海量数据流的变化,同时也减少了计算复杂度。通过仿真数据集进行实验分析,验证了提出的优化动态数据流断点区域检测方法在移动网络数据流离群数据检测挖掘方面的有效性,聚类检测的精度与效率具有明显优势。
Data flow breakpoint area detection is used to identify abnormal data in dynamic data flow. Because the mobile network data stream has the characteristics of infinite input and dynamic change,In view of existing algorithms,there are disadvantages of low clustering precision and poor efficiency of cluster points. It is difficult to realize real-time high precision detection of data flow breakpoints in mobile network environment.A dynamic data flow breakpoint optimization detection algorithm based on improved weighted neighbor density is proposed.The algorithm USES the density of microclusters to find outliers and cluster connections in data streams. To represent the evolution information of dynamic data flow,In order to reflect the original characteristics of data flow more accurately,The dynamic cluster update maintenance mechanism is introduced in the algorithm model. And by improving the right value proximity algorithm,When the computational model detects that a new object class is generated,Based on clustering algorithm,the model is reconstructed. So that you can find a huge amount of data flow,It also reduces computational complexity. Through the simulation data set for experimental analysis,The effectiveness of the proposed new dynamic data flow breakpoint detection method in the data detection and mining of mobile network data is verified. The accuracy and efficiency of detection have obvious advantages.
作者
吴振涛
闵俊
WU Zhen-tao;MIN Jun(School of Mathematics & Computer Science,Wuhan Polytechnic University,Wuhan Hubei 430023,China)
出处
《计算机仿真》
北大核心
2019年第1期470-474,共5页
Computer Simulation
关键词
动态数据流
断点区
微簇
改进权值邻近
聚类算法
Dynamic data flow
Breakpoint area
Micro cluster
Improved weight proximity
Clustering algorithms
