基于后验概率的有限偏序分类模型-概率论与数理统计专业论文.docxVIP

AbstractFinite Abstract Finite pantally ordered classification models are／isf!ful for rnany’sl atisl k：at rlppli cations imluding cognitive model．Finite partially ordeied set is a nattliaI mod(、j fbl cognitive problems，aS it is inore Ieasonable than others．Finite partially 01deied elassi fleaLion models are useful in many aspects，such as those in autolllated lnedical diagnosis or bioinfblmatics。Ollr main WOIk focuses On t,tle classification of patdaily OIdeled lTlOd (、ls A削at．islical framework is plesented for data analysis(If finil，(、parltall3’Ol(1¨f，d (iassilication models We use Bayesian method and post eri()1 I)I(／I)H1)jlit、’1{)II(、c1I¨l(、 classifieat．i《)11 tnoblem In response distribution pararoetel est,itnatiol／GibbK sampling procedure is ut．ilizl：d‘lhe psiamelcIs Lo be estimated arc Betnoulli response patsill(： ters We change tile ilni[1bed of tesl lug subject，S in t,he up—sel，ofto adapt the BoI noulli response parameters．Partially eldered model is an important tool to tleat this sit tla— tion Furthermore，an example in Tatsuoka f2002)、、rill be llSe(1 r．o ilhlstl ate(】1}1 1)10pos(、(1 procedure Our leseaich ls an extension of。the work done in sequential elassiikatj(Jll I)＼+【a1 suoka ＼Ve propose a ne＼,v method and compare this method with Shannon entrop3 pt oce,dlu e A method dealing with vague situation is presented in this papei Somet im(：s WC Wall{ to identify a student,’S abilit3’，for example，whether the studem has{he abi{i z、’()E nol、 If A student doesnll．COl rectly aliswei the question．we Call‘{．simply／h、Hal(、that Il(、 hasll’l{}tO abilil?，because nlaybe lie has lnade a careless mist ahs¨w(1 should He(’el¨I aspec{S If lhe classification is too st，riel we w(】nlt，gel，satisfaetorY result．【ll this pap(、I Kullback—Leil)lel infornladon is used t,o analyze t．he est，)trait．ion of para rTlt31．¨all()the Irl(／d(、1．Moreover．we bring forward a rlew idea of the sit，uation that one item“、st s sev(!l川abllities Key words：partially ordered set；up—set；Kullbaek—beibler inlb,n