KNN weighted reduced universum twin SVM for class imbalance learning

In real world problems, imbalance of data samples poses major challenge for the classification problems as a particular class are dominating. Problems like fault and disease detection involve hence need attention to avoid bias towards class. The models support vector machines (SVM) get biased majority results in misclassification minority samples. SVM suffers no prior information related is involved generation hyperplanes. Also, local neighbourhood ignored thus treats each sample equally generating However, points may be contaminated mislead Inspired by idea information, we propose K-nearest neighbour based weighted reduced universum twin learning (KWRUTSVM-CIL). proposed KWRUTSVM-CIL embodies uses balance classes problems. Local incorporated via weight matrix objective function. model, vectors used corresponding constraints functions exploit interclass information. oversampling undersampling approaches followed Universum gives data. Twin SVM, implement empirical risk minimization principle lead overfitting. model regularization term maximize margin structural which marrow statistical overcomes issues Experimental analysis signify that generalization ability superior comparison other models. As an application, use diagnosis Alzheimer’s breast cancer disease. showed better performance compared biomedical datasets.