计算机技术与发展
計算機技術與髮展
계산궤기술여발전
COMPUTER TECHNOLOGY AND DEVELOPMENT
2015年
4期
166-170,176
,共6页
陈运文%吴飞%吴庐山%刘博
陳運文%吳飛%吳廬山%劉博
진운문%오비%오려산%류박
时间序列%异常检测%数据挖掘%多元时间序列
時間序列%異常檢測%數據挖掘%多元時間序列
시간서렬%이상검측%수거알굴%다원시간서렬
time series%anomaly detection%data mining%multivariate time series
时间序列是一种重要类型的时态数据,广泛应用于科研、经济和军事等各个领域,而针对时间序列的异常检测研究是近年来大数据挖掘的热点与难点。文中以国内外最近的研究成果和时间序列检测的研究价值为基础,探讨了时间序列异常检测的定义并对相关异常检测方法进行归类研究与总结,同时指出每种异常检测方法的优缺点,并进一步分析部分具有代表性的时间序列异常检测的相关研究成果,尤其是讨论了多元时间序列异常检测研究所面临的难题,并给出解决此难题的思路和方法。最后总结归纳时间序列异常检测的几点建议与未来研究方向,以期对相关研究提供有益的参考。
時間序列是一種重要類型的時態數據,廣汎應用于科研、經濟和軍事等各箇領域,而針對時間序列的異常檢測研究是近年來大數據挖掘的熱點與難點。文中以國內外最近的研究成果和時間序列檢測的研究價值為基礎,探討瞭時間序列異常檢測的定義併對相關異常檢測方法進行歸類研究與總結,同時指齣每種異常檢測方法的優缺點,併進一步分析部分具有代錶性的時間序列異常檢測的相關研究成果,尤其是討論瞭多元時間序列異常檢測研究所麵臨的難題,併給齣解決此難題的思路和方法。最後總結歸納時間序列異常檢測的幾點建議與未來研究方嚮,以期對相關研究提供有益的參攷。
시간서렬시일충중요류형적시태수거,엄범응용우과연、경제화군사등각개영역,이침대시간서렬적이상검측연구시근년래대수거알굴적열점여난점。문중이국내외최근적연구성과화시간서렬검측적연구개치위기출,탐토료시간서렬이상검측적정의병대상관이상검측방법진행귀류연구여총결,동시지출매충이상검측방법적우결점,병진일보분석부분구유대표성적시간서렬이상검측적상관연구성과,우기시토론료다원시간서렬이상검측연구소면림적난제,병급출해결차난제적사로화방법。최후총결귀납시간서렬이상검측적궤점건의여미래연구방향,이기대상관연구제공유익적삼고。
Time series is a temporal data of important class which are widely used in scientific research,economics,and military and other fields,but the study aiming at anomaly detection of time series in recent years is the hot and difficult of data mining. In this paper,based on the recent research achievements at home and abroad and the research value of time sequential detection,discuss the definition of a-nomaly detection of time series and the research on related anomaly detection methods are classified and summarized,at the same time point out the advantages and disadvantages of each type of anomaly detection methods,and further analyze relevant research achievements of some typical time series of the anomaly detection,especially discussing the multivariate time series,the challenges faced by the institute of anomaly detection,and giving ideas and methods to solve this problem. Finally,some suggestions about anomaly detection and future research trends are also summarized,which is hopefully beneficial to the researchers of time series and other relative domains.