A framework for imprecise robust one-class classification models

摘要: A framework for constructing robust one-class classification models is proposed in the paper. It based on Walley’s imprecise extensions of contaminated which produce a set probability distributions data points instead single empirical distribution. The minimax and minimin strategies are used to choose an optimal distribution from construct separating functions. shown that algorithm computing parameters determined by extreme reduced finite number standard SVM tasks with weighted points. Important special cases models, including pari-mutuel, constant odd-ratio, Kolmogorov–Smirnov bounds studied. Experimental results synthetic real illustrate models.