基于统计特征的时序数据符号化算法.pdfVIP

31 10 V ol . 31 N o. 10 2008 10 CH IN ESE JO U RN A L O F CO M PU T ER S Oct . 200 8 钟清流 蔡自兴 ( 4 10083) SA X( ) , , SA X , , . SAX , SAX . . T P 18 The Symbolic Algorithm for Time Series Data Based on Statistic Feature ZH O N G Q ingL iu CA I Z iXing ( Center of Intell ig ent Sy stem and S of tw are , Sch ool of Inf or mat ion S cience and Engineering , Cent ral S ou th Univ ersity , Chang sha 4 10083) Abstract A new sym bo lic alg orit hm f or T im e Series Dat a Based on St at istic F eat ure is put fo r w ard in order t o surmount t he bugs w it h w hich SA X ( Symbolic A gg reg ate A ppro ximat ion) A lgo r it hm can no t describe tim e series inf ormat ion f ul ly . T his algor it hm, dif fering fro m the SA X, con sider ed t he sy mbolic as v ect or, and M ean and variance f rom each sub sequ ence w er e regarded as component s by w hich it s mean v alu e and radiat ion deg ree are described respect ively . Since it could provide more inf ormat ion described time ser ies th an SA X do , mo re accuracy result could be get w hen it is applied t o t im e series datamining. It s ex cellent behave. have been prov ed by a lot of exp er im ent s. Keywords time series analysis sym bolic representation sy mbolic aggregate approx im at ion ( SA X) , : 1 ( 1) , . ( 2) , , , . L in K eo gh [ 12] . 2003 !! ! , ( DF T ) [ 3] [ 8]