IND是一个开源的系统，可以处理大部分的独立事件，而这些独立事件都是使用固定长度的向量描述的值。IND提供了一系列的功能特征和使用风格，主要是为了方便普通用户以及高级用户或者是那些对调查研究感兴趣的人使用。IND是由四个基础的例程组成：数据操作例程、目录生成例程、目录检测例程和目录显示例程。
INDisapplicabletomostdatasetscosistigofidepedetistaces,eachdescribedbyafixedlegthvectorofattributevalues.Aattributevaluemaybeaumber,oeofasetofattributespecificsymbols,oromitted.Oeoftheattributesisdelegatedthe"target"adINDgrowstreestopredictthetarget.Predictiocathebedoeoewdataorthedecisiotreepritedoutforispectio.
INDprovidesarageoffeaturesadstyleswithcoveieceforthecasualuseraswellasfie-tuigfortheadvaceduserorthoseiterestediresearch.INDcabeoperatediaBreima/Friedma/Olshe/Stoe-likemode(butwithoutregressiotrees,surrogatesplitsormultivariatesplits),adiamodelikeC4.5.Advacedfeaturesallowmoreextesivesearch,iteractivecotroladdisplayoftreegrowig,adBayesiaadMMLalgorithmsfortreepruigadsmoothig.Theseofteproducemoreaccurateclassprobabilityestimatesattheleaves.
INDalsocomeswithacomprehesiveexperimetalcotrolsuite.INDcosistoffourbasickidsofrouties;datamaipulatiorouties,treegeeratiorouties,treetestigrouties,adtreedisplayrouties.Thedatamaipulatioroutiesareusedtopartitioasiglelargedatasetitosmallertraiigadtestsets.Thegeeratioroutiesareusedtobuildclassifiers.Thetestroutiesareusedtoevaluateclassifiersadtoclassifydatausigaclassifier.Adthedisplayroutiesareusedtodisplayclassifiersivariousformats.
INDiswritteiK&RC,withcotrolligscriptsithe"csh"shellofUNIX,adextesiveUNIXmaetries.ItisdesigedtobeusedoayUNIXsystem,althoughithasolybeethoroughlytestedoSUNplatforms.INDcomeswithamaualgivigaguidetotreemethods,adpoiterstotheliterature,adseveralcompaiodocumets.
评论