LAUSR

Breast cancer is the second leading cause of cancer related death for women in the world, and mass is one of the most common kinds of abnormal. A mass can be either benign or malignant, the accurate diagnosis is important for early intervention and treatment. In this paper, we investigated the mass classification problem and proposed a new method for feature selection. The proposed method integrates joint $${l_{2,1}}$$l2,1-norm minimizing regularization with a nonparallel twin support vector machine, which is called TWSVML21. The $${l_{2,1}}$$l2,1-norm regularization selects features across positive and negative classes with joint sparsity, and features are selected by a ranking strategy. An iterative method is proposed to solve the involved optimization problem. Preliminary results on mass classification and several benchmark datasets showed the feasibility and effectiveness of the proposed TWSVML21 method.