Search

Classification is one of the most important topics in machine learning. However, most of these works focus on the two-class classification (i.e., classification into 'positive' and 'negative'), whereas studies on multi-class classification are far from enough. In this study, we develop a novel methodology of multiple classifier systems (MCS) with one-vs-one (OVO) scheme for the multi-class classification task. First, the multi-class classification problem is divided into as many pairs of easier-to-solve binary sub-problems as possible. Subsequently, an optimal MCS is generated for each sub-problem using a roulette-based feature subspace selection and validation procedure. Finally, to identify the final class of a query sample, an OVO aggregation strategy is employed to obtain the class from the confidence score matrix derived from the MCS. To verify the effectiveness and robustness of the proposed approach, a thorough experimental study is performed. The extracted findings supported by the proper statistical analysis indicate the strength of the proposed method with respect to the state-of-the-art methods for multi-class classification problems. [ABSTRACT FROM AUTHOR]

The One-vs-One (OVO) scheme that decomposes the original more complicated problem into as many as possible pairs of easier-to-solve binary sub-problems is one of the most popular techniques for handling multi-class classification problems. In this paper, we propose an improved Dynamic Ensemble Selection (DES) procedure, which aims to enhance the OVO scheme via dynamically selecting a group of appropriate heterogeneous classifiers in each sub-problem for each query example. To do so, twenty heterogeneous classification algorithms are selected to obtain a set of candidate classifiers for each sub-problem derived from the OVO decomposition. Then, a simple yet efficient DES procedure is developed to execute the dynamic selection for each query example in each sub-problem. Finally, all the selected binary heterogeneous ensembles are combined by using majority voting to obtain the final output class. To evaluate the proposed method, we carry out a series of experiments on twenty datasets selected from the KEEL repository. The results supported by proper statistical tests demonstrate the validity and effectiveness of our proposed method, compared with state-of-the-art methods for OVO-based multi-class classification. [ABSTRACT FROM AUTHOR]

Multi-class classification problems can be addressed by using decomposition strategy. One of the most popular decomposition techniques is the One-vs-One (OVO) strategy, which consists of dividing multi-class classification problems into as many as possible pairs of easier-to-solve binary sub-problems. To discuss the presence of classes with different cost, in this paper, we examine the behavior of an ensemble of Cost-Sensitive Back-Propagation Neural Networks (CSBPNN) with OVO binarization techniques for multi-class problems. To implement this, the original multi-class cost-sensitive problem is decomposed into as many sub-problems as possible pairs of classes and each sub-problem is learnt in an independent manner using CSBPNN. Then a combination method is used to aggregate the binary cost-sensitive classifiers. To verify the synergy of the binarization technique and CSBPNN for multi-class cost-sensitive problems, we carry out a thorough experimental study. Specifically, we first develop the study to check the effectiveness of the OVO strategy for multi-class cost-sensitive learning problems. Then, we develop a comparison of several well-known aggregation strategies in our scenario. Finally, we explore whether further improvement can be achieved by using the management of non-competent classifiers. The experimental study is performed with three types of cost matrices and proper statistical analysis is employed to extract the meaningful findings. [ABSTRACT FROM AUTHOR]