Accuracy and diversity in classifier selection for one-class classification ensembles

One-class classification is one of the most challenging topics in the field of machine learning. Recently creating Multiple Classifier Systems for this task has proven itself as a promising research direction. Here arises a problem on how to select valuable members to the committee - so far a largely unexplored area in one-class classification. This paper introduced a novel approach that allows to choose appropriate models to the committee in such a way that assures both high accuracy of individual models and a high diversity among the pool members. We aim at preventing the selection of both too weak or too similar models. This is achieved with the usage of a multi-objective optimization that allows for more than one criterion while searching for a good combination of predictors. A memetic algorithm is applied due to its efficiency and less random behavior than traditional genetic algorithm. Used diversity measures are designed specially to fully exploit the specific nature of the one-class classifiers under consideration. Experimental results carried on a number of benchmark datasets proves the quality of our method and that it outperforms traditional single-objective approaches.