Interclass Interference Suppression in Multi-Class Problems

Multi-classifiers are widely applied in many practical problems. But the features that can significantly discriminate a certain class from others often deleted feature selection process of multi-classifiers, which seriously decreases generalization ability. This paper refers to this phenomenon as interclass interference multi-class problems and analyzes its reason detail. Then, summarizes three suppression methods including method based on all-features, one-class classifiers binary compares their effects via 10-fold cross-validation experiments 14 UCI datasets. Experiments show suppress efficiently obtain best classification accuracy among methods. Further were done compare effect two one-versus-one one-versus-all method. Results better accuracy. By proposing concept inference studying methods, improves ability multi-classifiers.