The Application of Weighted Kernel Fisher Discriminant Analysis in Student Loans Default
This paper took 2782 data of non-defaulted and defaulted state-subsidized student loan in a university as samples. Firstly, by using Factor Analysis, 7 factors were picked up from original 12 attributes of every sample. Then 70％ data were served as training samples and 30％ data were served as test samples. Furthermore, Fisher Discriminated, Bayesian Discriminate and Weighted Kernel Fisher Discriminated were respectively used to classify these data. The result indicated that the accuracy rate of Fisher Discriminated was 54.08％, while the accuracy rate of Bayesian was 67.99％ and Weighted Kernel Fisher Discriminated reached 74.0％. To decision and management, this research has guiding significance for banks, and the principle and the method can also be applied into other similar problems.
Factor analysis Weighted kernel fisher discriminated analysis Loan defaults
Tang Qin Zeng Jianyou Li Xing Zhang Hongyang
School of Mathematics and Physics, China University of Geosciences, Wuhan, P. R. China, 430074 School of Arts and Communication, China University of Geosciences, Wuhan, P. R. China, 430074
国际会议
The 8th International Conference on Innovation and Management(第八届创新与管理国际会议 ICIM 2011)
日本福冈
英文
1274-1277
2011-11-30（万方平台首次上网日期，不代表论文的发表时间）