基于Apache AsterixDB的相似性查询-豆柴文库

您所在位置：网站首页 / 基于Apache AsterixDB的相似性查询.docx / 文档详情

在线预览结束，喜欢就下载吧，查找使用更方便

5 金币

下载文档

/ 2

下载提示

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

文本预览

基于ApacheAsterixDB的相似性查询
Introduction
ApacheAsterixDBisahighlyscalable,opensourceBigDataManagementPlatformthatcombinesdifferentdatamodelssuchassemi-structured,structured,andunstructureddatainreal-time.TheplatformleveragesastoragelayerthatisbasedontheConcurrentAppend-OnlyB-Tree(CAO-B-Tree)toprovidehighperformanceandscalabilityforbigdataapplications.Additionally,AsterixDBprovidesaflexibleandpowerfulquerylanguagecalledAQL(AsterixQueryLanguage)tosupportcomplexanalyticalqueries.OneusefulandpowerfultypeofquerythatcanbeperformedinAsterixDBissimilaritysearch.
Similaritysearchreferstotheprocessofidentifyingdatapointsthataresimilartoagivenquerypoint.Thistypeofqueryisessentialinmanyapplicationssuchasrecommendationsystems,anomalydetection,andfrauddetection.Inthispaper,wewilldiscusstheimplementationofsimilaritysearchusingAsterixDB.
Background
Similaritysearchalgorithmsarecharacterizedbythetypeofdistancefunctionusedtomeasurethesimilaritybetweendatapoints.ThemostcommonlyuseddistancefunctionsincludeEuclideandistance,Manhattandistance,cosinesimilarity,andJaccarddistance.Inadditiontothedistancefunction,similaritysearchalgorithmscanbedividedintotwocategories:exactandapproximatesearch.Exactsearchalgorithmsprovideexactmatchestothequerypoint,whileapproximatesearchalgorithmsprovideanapproximatesetofmatcheswithinacertainthreshold.
Traditionalapproachestosimilaritysearchincludek-nearestneighbors(k-NN)searchandrangesearch.k-NNsearchinvolvesfindingthekdatapointsthatareclosesttothequerypoint.Rangesearchinvolvesfindingalldatapointswithinacertainradiusofthequerypoint.
Implementation
AsterixDBprovidesseveralbuilt-infunctionsforsimilaritysearch,includingcosinesimilarityandJaccarddistance.Thesefunctionscanbeusedincomplexanalyticalqueriestoidentifysimilardatapoints.
ToperformsimilaritysearchinAsterixDB,wefirstneedtodefinethesimilarityfunctionandthequerypoint.Thesimilarityfunctiontakestwoinputdatapointsandreturnsasimilarityscorebasedonthedistancebetweenthem.Forexample,thecosinesimilarityfunctiontakest