Title :
An improved multi-class SVM algorithm and its application to the credit scoring model
Author :
Tian, Xiang ; Deng, Feiqi
Author_Institution :
Coll. of Autom. Sci. & Eng., South China Univ. of Technol., Guangzhou, China
Abstract :
On the basis of the traditional SVM principle and a "one-by-one" classifier constructing strategy, a new multi-class SVM, named Binary Tree Multistage Support Vector Machine (BTMSVM), is proposed. This classifier is simple and results in less duplicating training samples. The credits of 96 listed companies of China in 2000 are evaluated with this SVM credit scoring model, and the simulation results show that a high classification accurate rate of up to 98.11% is attained.
Keywords :
credit transactions; statistical analysis; support vector machines; trees (mathematics); binary tree multistage support vector machine; credit scoring model; linear classification accuracy; statistical analysis; training samples; Binary trees; Classification tree analysis; Educational institutions; Mathematical model; Mathematics; Support vector machine classification; Support vector machines;
Conference_Titel :
Intelligent Control and Automation, 2004. WCICA 2004. Fifth World Congress on
Print_ISBN :
0-7803-8273-0
DOI :
10.1109/WCICA.2004.1341918
