決策樹生成原理

決策樹生成原理

ID:41933804

大?。?7.00 KB

頁數(shù):5頁

時間:2019-09-04

決策樹生成原理_第1頁
決策樹生成原理_第2頁
決策樹生成原理_第3頁
決策樹生成原理_第4頁
決策樹生成原理_第5頁
資源描述:

《決策樹生成原理》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在工程資料-天天文庫

1、決策樹生成原理AbstractThispaperdetailstheID3classificationalgorithm.Verysimply,ID3buildsadecisiontreefromafixedsetofexamples.Theresultingtreeisusedtoclassifyfuturesamples.Theexamplehasseveralattributesandbelongstoaclass(likeyesorno).Theleafnodesofthedecisiont

2、reecontaintheclassnamewhereasanon-leafnodeisadecisionnode?Thedecisionnodeisanattributetestwitheachbranch(toanotherdecisiontree)beingapossiblevalueoftheattribute.ID3usesinformationgaintohelpitdecidewhichattributegoesintoadecisionnode?Theadvantageoflearn

3、ingadecisiontreeisthataprogram,ratherthanaknowledgeengineer,elicitsknowledgefromanexpert?IntroductionJ.RossQuinlanoriginallydevelopedID3attheUniversityofSydney?HefirstpresentedID3in1975inabook,MachineLearning,vol.1,no.1.ID3isbasedofftheConceptLearningS

4、ystem(CLS)algorithm.ThebasicCLSalgorithmoverasetoftraininginstancesC:Step1:IfallinstancesinCarcpositive,thencreateYESnodeandhalt.IfallinstancesinCarenegative,createaNOnodeandhalt.Otherwiseselectafeature,Fwithvaluesvl,vnandcreateadecisionnode?Step2:Part

5、itionthetraininginstancesinCintosubsetsCl,C2,…,CnaccordingtothevaluesofV.Step3:applythealgorithmrecursivelytoeachofthesetsCi.Note,thetrainer(theexpert)decideswhichfeaturetoselect.ID3improvesonCLSbyaddingafeatureselectionheuristic?ID3searchesthroughthea

6、ttributesofthetraininginstancesandextractstheattributethatbestseparatesthegivenexamples?IftheattributeperfectlyclassifiesthetrainingsetsthenID3stops;otherwiseitrecursivelyoperatesonthen(wheren=numberofpossiblevaluesofanattribute)partitionedsubsetstoget

7、their”best”attribute.Thealgorithmusesagreedysearch,thatis,itpicksthebestattributeandneverlooksbacktoreconsiderearlierchoices?DiscussionID3isanonincrementalalgorithm,meaningitderivesitsclassesfromafixedsetoftraininginstances-Anincrementalalgorithmrevise

8、sthecurrentconceptdefinition,ifnecessary,withanewsample?Theclassescreatedby1D3areinductive,thatis,givenasmallsetoftraininginstances,thespecificclassescreatedbyID3areexpectedtoworkforallfutureinstances?Thedistributionoftheunknownsmustbet

當(dāng)前文檔最多預(yù)覽五頁,下載文檔查看全文

此文檔下載收益歸作者所有

當(dāng)前文檔最多預(yù)覽五頁,下載文檔查看全文
溫馨提示:
1. 部分包含數(shù)學(xué)公式或PPT動畫的文件,查看預(yù)覽時可能會顯示錯亂或異常,文件下載后無此問題,請放心下載。
2. 本文檔由用戶上傳,版權(quán)歸屬用戶,天天文庫負(fù)責(zé)整理代發(fā)布。如果您對本文檔版權(quán)有爭議請及時聯(lián)系客服。
3. 下載前請仔細(xì)閱讀文檔內(nèi)容,確認(rèn)文檔內(nèi)容符合您的需求后進行下載,若出現(xiàn)內(nèi)容與標(biāo)題不符可向本站投訴處理。
4. 下載文檔時可能由于網(wǎng)絡(luò)波動等原因無法下載或下載錯誤,付費完成后未能成功下載的用戶請聯(lián)系客服處理。