基于RCSW的数据流速度异常检测算法研究  

An Algorithm for Data Stream Speed Anomaly Detection Based on RCSW

作　　者：詹英[1] 吴春明[2] 王宝军[1] 

机构地区：[1]浙江交通职业技术学院,浙江杭州311112 [2]浙江大学系统工程研究所,浙江杭州310027

出　　处：《电子学报》2012年第4期674-680,共7页Acta Electronica Sinica

基　　金：浙江省自然科学基金(No.Y1101133);浙江省教育厅项目(No.Y201018946)

摘　　要：目前许多应用领域产生数据流的流速不断地震荡,使得面向数据流的挖掘变得困难.系统采用RCSW来完成数据流抽取,提出了实时度T、关键时点集、数据流处理率的概念,并进一步提出了数据流速度异常检测算法.系统监控、预测数据流速,当数据流速异常减速或增速时,系统智能调节环形缓冲区和数据流处理率来应对异常,为解决数据流处理能力与流速、流量与有限空间之间的矛盾提供解决方案.实验表明数据流速度异常检测算法能够保证数据流的挖掘持续正常实施,最大程度的满足系统的实时性要求.In many fields,data stream continues to grow in terms of generation speed,scale and vibration,which makes the data stream mining more difficult.RCSW is used in data stream mining to finish data steam sampling.The three concept such as real-time T,key time point set,data stream processing ratio are proposed.Then an algorithm for data stream speed anomaly detection is proposed,which monitor and predict flow velocity.The system intelligently adjust ring buffer and data stream processing ratio if there is excessive flow velocity,in order to solve the conflicts commendably between data processing power and flow velocity,throughput and limited memory.Experimental results show that it is an algorithm for data stream speed anomaly detection which can ensure normal execution of data stream mining and well meet the need of the system real-time.

关 键 词：数据流 环形循环滑动窗口 关键时点 实时度T 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]