Bagging schemes on the presence of class noise in classification

A biclustering approach for classification with mislabeled data

“…Other approaches are variants of the Bagging and Boosting techniques (Abellán & Masegosa, 2012;Cantador & Dorronsoro, 2005;Cao, Kwong, & Wang, 2012). Here, the same classifier is trained by using different samples from the dataset, leading to different class boundaries that are combined afterwards.…”

“…Here, the same classifier is trained by using different samples from the dataset, leading to different class boundaries that are combined afterwards. The main argument behind the Bagging schemes devised by Abellán and Masegosa (2012) in particular is that it is expected that each mislabeled instance will be part of just a few of the generated training sets, thus a majority of the boundaries will not be influenced by such instance. Even though the proposed schemes usually improve the performance, their success still strongly depends on which instances were mislabeled.…”

“…These methods are usually referred to as data cleansing methods (Brodley & Friedl, 1999;Guan, Yuan, Lee, & Lee, 2011;Zhu, Wu, & Chen, 2006). On the other hand, there are approaches, called noise-tolerant (or noise-robust), that can deal intrinsically with label noise while inducing the classifier models and, thus, do not depend on data preprocessing (Abellán & Moral, 2003;Abellán & Masegosa, 2012;Bootkrajang & Kabán, 2012. Finally, there are also some approaches, referred to here as hybrid ones, which combine features and properties of the abovementioned classes, e.g.…”

Analysis of Credal-C4.5 for classification in noisy domains

“…Besides, its use with bagging ensemble ( Abellán & Mantas, 2014;Abellán & Masegosa, 2009a;2012a ) and its above mentioned extension ( Mantas & Abellán, 2014a ) are especially suitable when noisy data are classified. A complete and recent revision of machine learning methods to manipulate label noise can be found in Frenay and Verleysen (2014) .…”

A New Robust Classifier on Noise Domains: Bagging of Credal C4.5 Trees

“…This model obtains improvements with respect to other known ensembles of classifiers used in this type of setting: the bagging scheme with the C4.5 model and the known classifier Random Forest (RF). It is shown in the literature that the bagging scheme with the C4.5 model is normally the winning model in many studies about classification noise [23,24].…”