Angle-Based Distance-Weighted Support Vector Machine in Multicategory Classification

• Activity Number: 445
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Hui Sun* and Bruce A. Craig and Lingsong Zhang
• Companies: Purdue University and Purdue University and Purdue University
• Keywords: Discriminant analysis ; Imbalanced data ; High dimension ; Support vector machine

Abstract

Classification is an important supervised learning technique with numerous applications. We proposed an angle-based multicategory classification based on binary Distance-weighted Support Vector Machine (DWSVM) classification. The new method has the merits of both Support vector machine (SVM) and Distance weighted discrimination (DWD). It alleviate both the data pilling issue of SVM and the imbalanced data issue that DWD have by adopting the intercept from SVM. Using angle-based method, it considers a simplex based prediction rule which rules out the sum to zero constraint. Theoretical and Numerical studies demonstrate the advantage of this new angle-based DWSVM method.