Forming Ensembles of Soft One-Class Classifiers with Weighted Bagging

Abstract

For many real-life problems obtaining representative examples from a given class is relatively easy, while for the remaining ones are difficult, or even impossible. However, we would still like to construct a pattern classifier that could distinguish between the known and unknown cases. In such cases we are dealing with one-class classification, or learning in the absence of counterexamples. Such recognition systems must display a high robustness to new, unseen objects that may belong to an unknown class. That is why ensemble learning has become an attractive perspective in this field. In our work, we propose a novel one-class ensemble classifier, based on weighted Bagging. Wagging method is used to obtain randomized weights and utilize them directly in the process of training Weighted One-Class Support Vector Machines. This introduces a diversity into the pool of one-class classifiers and extends the competence of formed ensemble. Additionally, to discard similar or weak classifiers we propose to add a clustering-based pruning procedure to our ensemble. It works on the basis of similarities between weights used by each base model and detecting groups of similar predictors. This allows us to reduce the number of classifiers in the pool by selecting a single representative for each cluster. Experimental analysis, carried out on a number of benchmarks and backed-up with statistical analysis proves that the proposed method can outperform state-of-the-art ensembles dedicated to one-class classification.

Computational Biology and Machine Learning Approaches to Understand Mechanistic Microbiome-Host Interactions

“…Ensemble learning ( Che et al, 2011 ), combines multiple individual classifiers to achieve a final classification and has been used to predict PPI based HIV-human and hepatitis C virus-human networks ( Mei, 2013 ; Emamjomeh et al, 2014 ). Ensemble classification methods outperform individual classifiers based on several use-cases ( Krawczyk, 2015 ; Haque et al, 2016 ; Yijing et al, 2016 ; Lin et al, 2019 ) and can be generalized into three distinct categories namely bagging, boosting and stacked generalization. The last of the three approaches, stacked generalization, was used by Emamjomeh et al (2014) to predict PPIs between human and the hepatitis C virus.…”

An Ensemble Classification Method for High‐Dimensional Data Using Neighborhood Rough Set

“…Zhang and Cao proposed an ensemble pruning method based on spectral clustering [22]. Krawczyk used a cluster-based pruning method in the weighted Bagging ensemble classification [23]. However, these methods do not consider the diversity among base classifiers and the classification performance of the classifiers at the same time when calculating the distance between base classifiers.…”