摘要
总结目前数据流在线检测算法的优缺点,提出了一种新的数据流在线检测算法—SWKLOF。该算法采用滑动时间窗口对数据流进行封装,用k-距离进行剪枝,剔除大部分正常数据,对剩余疑似异常数据采用局部离群因子LOF(local outlier factor)进一步精确筛选。理论分析和实验结果表明该算法降低了时间复杂度,提高了检测准确性。
Summarizing the advantages and disadvantages of the current datastreams clustering algorithms,this paper proposes a new datastreams clustering algorithm——SWKLOF. The algorithm uses sliding time window to encapsulate the datastreams,using the k-distance to pruning,then removing most of the normal data. At last,using the local outlier factor to screen accurately the remaining data which are suspected abnormal datas. Theoretical analysis and experimental results show that the algorithm reduces the time complexity and improves greatly the accuracy of datastreams detection.
出处
《科学技术与工程》
北大核心
2014年第34期219-223,共5页
Science Technology and Engineering
基金
"十二五"国家科技支撑计划项目(2012BAF12B14)
贵州省重大科技专项(黔科合重大专项字(2012)6018)
贵州省重大科技专项(黔科合重大专项字(2013)6019)资助
贵州省工业攻关项目(黔科合GY字(2013)3020)
