Affiliation:
1. Central University of Finance and Economics
Abstract
Abstract
This study extends the proximal support vector machine (PSVM) to the ordinal regression problems by proposing a proximal ordinal support vector regression (POSVR) based on the fixed margin strategy. The proposed POSVR is applied to the multi-class classification problems with ordered discrete outcome events under the assumption that the ordered classes can be separated by a group of parallel decision hyperplanes. POSVR can be transformed into a linear equation system and solved conveniently. The experimental results on the benchmark datasets and a real-world recovery rate dataset of US corporate bonds demonstrate that POSVR achieves better performance in terms of cross-validation outputs on the training set and out-of-sample prediction on the testing set. It implies that the proposed POSVR is a promising technique for credit risk modeling and other areas in ordinal regression problems.
Publisher
Research Square Platform LLC
Reference44 articles.
1. Comparative study among three strategies of incorporating spatial structures to ordinal image regression;Tian Q;Neurocomputing,2014
2. Cost-sensitive ordinal regression for fully automatic facial beauty assessment;Yan H;Neurocomputing,2014
3. Computer aided detection for fibrillations and flutters using deep convolutional neural network;Fujita H;Inform Sci,2019
4. Chang K-Y, Chen C-S (2011) & Y-P. Hung Ordinal hyperplanes ranker with cost sensitivities for age estimation, In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 585–592
5. Support vector machines in ordinal classification: An application to corporate credit scoring;Dikkers H;Neural Netw World,2005