RUS  ENG
Full version
JOURNALS // Informatika i Ee Primeneniya [Informatics and its Applications] // Archive

Inform. Primen., 2013 Volume 7, Issue 4, Pages 105–111 (Mi ia290)

This article is cited in 2 papers

Prediction and classification method for censored data

T. V. Zakharova, E. M. Abramova

Department of Mathematical Statistics, Faculty of Computational Mathematics and Cybernetics, M. V. Lomonosov Moscow State University, Moscow 119991, Russian Federation

Abstract: The classification method for noninsulin-dependent diabetes mellitus patients cohort is presented and the technique for identification of diabetes mellitus indicators is described. The basic medical data we dealt with turned out to be unfit for classification. The main obstacle for applying classical discrimination approaches was insufficiency and incompleteness of original data. For data processing, the authors suggest to select different sets of discriminant characteristics and to obtain classification functions for each set. The number of these sets depends on data incompleteness degree. The more data are omitted, the more different sets are needed. Each patient finally refers to the group, for which he gets the greater number of matches in classification. This multistep procedure reimburses small sample size and insufficiency and incompleteness of original data.

Keywords: hypothesis; censored data; discriminant variables; classification functions; forecasting.

Received: 10.01.2013

DOI: 10.14357/19922264130411



Bibliographic databases:


© Steklov Math. Inst. of RAS, 2025