Search

This paper formulates a support vector machine with quantile hyper-spheres (QHSVM) for pattern classification. The idea of QHSVM is to build two quantile hyper-spheres with the same center for positive or negative training samples. Every quantile hyper-sphere is constructed by using pinball loss instead of hinge loss, which makes the new classification model be insensitive to noise, especially the feature noise around the decision boundary. Moreover, the robustness and generalization of QHSVM are strengthened through maximizing the margin between two quantile hyper-spheres, maximizing the inner-class clustering of samples and optimizing the independent quadratic programming for a target class. Besides that, this paper proposes a novel local center-based density estimation method. Based on it, ρ-QHSVM with surrounding and clustering samples is given. Under the premise of high accuracy, the execution speed of ρ-QHSVM can be adjusted. The experimental results in artificial, benchmark and strip steel surface defects datasets show that the QHSVM model has distinct advantages in accuracy and the ρ-QHSVM model is fit for large-scale datasets. [ABSTRACT FROM AUTHOR]

For strip steel surface defect dataset with abnormal samples and weakly relevant features, a novel multi-class classification method is realized in this paper. Firstly, weight parameter representing the importance degree of sample is obtained by calculating the clustering information of samples in the training dataset. Then, weight parameter representing the importance degree of feature is obtained by utilizing sparse important samples dataset and Relief-F algorithm. After that, a multiple support vector hyper-sphere with feature and sample weights (FSW-MSVH) is proposed. The novel classifier can realize multi-class classification. Finally, the multi-class classifier is used to do classification experiments for eight classes of strip steel surface defect. Experimental results show that FSW-MSVH classifier can greatly improve classification accuracy and speed. [ABSTRACT FROM AUTHOR]