Angle-based cost-sensitive multicategory classification

Many real-world classification problems come with costs which can vary for different types of misclassification. It is thus important to develop cost-sensitive classifiers minimize the total misclassification cost. Although binary have been well-studied, solving multicategory still challenging. A popular approach address this issue construct K functions a K-class problem and remove redundancy by imposing sum-to-zero constraint. However, such method usually results in higher computational complexity inefficient algorithms. In article, we propose novel angle-based framework without Loss that included are further justified be Fisher consistent. To show usefulness framework, two boosting algorithms derived as concrete instances. Numerical experiments demonstrate proposed yield competitive performances against other existing approaches.