A Novel Algorithm Model for Multi-class Classification

Abstract

Multi-class classification is an important and on-going research subject in machine learning. In this paper, we propose an algorithm model for k-class multi-class classification problem based on p-class (2≤ p ≤ k) support vector ordinal regression machine (SVORM). A series of algorithms can be generated by selecting the different parameters p, L and the code matrix. When p = 2, they reduce to the popular algorithms based on 2-class SVMs. When p = 3, they improve K-SVCR in [1] and ν-K-SVCR in [19]. The algorithms based on p- class SVORM in this algorithm model are more interesting because our preliminary numerical experiments show that then are promising. At last, some problems for further study are suggested. Key words: Multi-class classification problem; decomposition-reconstruction; support vector ordinal regression machine; error-correcting output code This work is supported by the Key Project of National Natural Science Foundation of China (No.10631070), the National Natural Science Foundation of China (No.10801112,No.70601033) and the China Postdoctoral Science Foundation funded project(No.20080430573)