. "0"^^ . . . . . "Projekt splnil vyt\u010Den\u00E9 c\u00EDle spo\u010D\u00EDvaj\u00EDc\u00ED v rozvoji metod, kter\u00E9 umo\u017En\u00ED popsat a porovnat v\u00FDkonnost diskrimina\u010Dn\u00EDch model\u016F.Byly odvozeny origin\u00E1ln\u00ED metody odhadu ROC\u00A0 pro modely s podm\u00EDn\u011Bn\u00FDm elipticko vrstevnicov\u00FDm rozd\u011Blen\u00EDm (bi-ECD modely). Sestaven\u00E9 pro"@cs . . "Neuvedeno."@en . . . . "GA402/04/1308" . . . "Classification models and the assessment of their predictive properties"@en . . . "39"^^ . . . "2007-10-16+02:00"^^ . "Discrimination belongs among decision support techniques that are frequently used at different application fields (bank clients scoring, repayment quality prediction, market segmentation etc.). The automation of the process of building discrimination models brings new questions in the area of comparison of different models when conditions of discrimination can change and are not known precisely. This research is focused on the field of analysis quality of discrimination models and on the estimation and on the use of the receiver operating characteristic (ROC) curve and area under curve (AUC). Research purpose covers different levels: studying the properties of discrimination models, systematic arrangement and overview of the research results acquired from the literature dealing with ROC and AUC, deeper studying and developing of the possibilities of kernel estimates of ROC, and studying the possibilities of how to describe and compare the performance of different types of classification"@en . "There were achieved the objectives of the project which was focused on development of methods for description and comparison the performance of classification methods.\u00A0\u00A0ROC curves were uniquely described for general fully parameterized bi-distributional"@en . "39"^^ . "http://www.isvav.cz/projectDetail.do?rowId=GA402/04/1308"^^ . . "1"^^ . "Klasifika\u010Dn\u00ED modely a porovn\u00E1n\u00ED jejich prediktivn\u00EDch vlastnost\u00ED" . . "1"^^ . "Diskriminace n\u00E1le\u017E\u00ED mezi metody pro podporu rozhodov\u00E1n\u00ED, kter\u00E9 maj\u00ED \u010Dast\u00E9 vyu\u017Eit\u00ED v praktick\u00FDch aplikac\u00EDch r\u016Fzn\u00E9ho zam\u011B\u0159en\u00ED (kategorizace klient\u016F banky, predikce kvality spl\u00E1cen\u00ED \u00FAv\u011Bru, segmentace z\u00E1kazn\u00EDk\u016F na trhu apod.)- Automatizace procesu sestavov\u00E1n\u00ED klasifika\u010Dn\u00EDch model\u016F p\u0159in\u00E1\u0161\u00ED nov\u00E9 probl\u00E9my, kter\u00E9 se t\u00FDkaj\u00ED ot\u00E1zek jak porovnat navz\u00E1jem r\u016Fzn\u00E9 modely, jestli\u017Ee podm\u00EDnky, p\u0159i kter\u00FDch se klasifikace prov\u00E1d\u00ED, jsou prom\u011Bnliv\u00E9 a nejsou p\u0159esn\u011B zn\u00E1m\u00E9. Pl\u00E1novan\u00FD v\u00FDzkum je orientov\u00E1n do oblasti anal\u00FDzy kvality diskrimina\u010Dn\u00EDch model\u016F a do oblasti odhadu a vyu\u017Eit\u00ED opera\u010Dn\u00ED charakteristiky modelu (ROC k\u0159ivky) a plochy AUC pod ROC k\u0159ivkou. Z\u00E1m\u011Br v\u00FDzkumu pokr\u00FDv\u00E1 r\u016Fzn\u00E9 \u00FArovn\u011B zkoum\u00E1n\u00ED: vlastnosti model\u016F diskriminace, systematick\u00E9 uspo\u0159\u00E1d\u00E1n\u00ED a p\u0159ehledliter\u00E1rn\u00EDch v\u00FDsledk\u016F z oblasti ROC a AUC, hlub\u0161\u00ED studium a dal\u0161\u00ED rozvoj vyu\u017Eit\u00ED metody j\u00E1drov\u00E9ho odhadu ROC k\u0159ivky a studium mo\u017Enost\u00ED, jak popsat a navz\u00E1jem porovnat v\u00FDkonnost r\u016Fzn\u00FDch typ\u016F diskrimina\u010Dn\u00EDch model\u016F (statistick\u00FDch a nestatistick\u00FDch)." . . .