A classification model based on incomplete information on features in the form of their average values
Lev Vladimirovich Utkin, Yu. A. Zhuk, Ivan A. Selikhovkin · Scientific and Technical Information Processing · 2012
This paper presents a model of classification under incomplete information in the form of mathematical expectations of features; it is based on the minimax (minimin) strategy of decision making. The discriminant function is calculated by maximization (minimization) of the risk functional as a measure of misclassification, by a set of distributions of probabilities with bounds determined by information on features, and minimization by the set of parameters. The algorithm is reduced to solution of the parametric problem of linear programming.