A high-dimensionality-trait-driven learning paradigm for high dimensional credit classification

Financial Innovation

Table 4 Robustness analysis of classification with PCA in China Unionpay dataset when 100 < feature dimensions < sample size

	Single linear classifier					Single nonlinear classifier
	Total	TP	TN	AUC		Total	TP	TN	AUC
Results of 10 experiments	0.7077	0.7512	0.6646	0.7593	Results of 10 experiments	0.6638	0.7274	0.6009	0.7019
	0.6883	0.7433	0.6344	0.7407		0.6564	0.6864	0.6271	0.6911
	0.7030	0.7286	0.6746	0.7420		0.6693	0.6895	0.6467	0.7038
	0.6879	0.7516	0.6250	0.7458		0.6658	0.7219	0.6103	0.7093
	0.6933	0.7431	0.6420	0.7502		0.6665	0.7030	0.6289	0.7047
	0.6960	0.7348	0.6574	0.7548		0.6660	0.6983	0.6337	0.7108
	0.6972	0.7461	0.6492	0.7397		0.6508	0.6889	0.6135	0.6830
	0.7108	0.7857	0.6375	0.7653		0.6636	0.7170	0.6114	0.7024
	0.6960	0.7329	0.6592	0.7427		0.6617	0.7026	0.6207	0.6976
	0.6933	0.7433	0.6444	0.7385		0.6574	0.7021	0.6137	0.6923
Average	0.6974	0.7461	0.6488	0.7479	Average	0.6621	0.7037	0.6207	0.6997
Standard deviation	0.0077	0.0158	0.0151	0.0092	Standard deviation	0.0056	0.0142	0.0135	0.0087

	Linear ensemble classifier					Nonlinear ensemble classifier
	Total	TP	TN	AUC		Total	TP	TN	AUC
Results of 10 experiments	0.6921	0.7017	0.6819	0.7624	Results of 10 experiments	0.6595	0.7148	0.6004	0.7054
	0.6883	0.7082	0.6689	0.7516		0.6593	0.7155	0.6043	0.7089
	0.6919	0.6947	0.6890	0.7551		0.6570	0.7159	0.5956	0.7004
	0.6479	0.6704	0.6246	0.6969		0.6077	0.7746	0.4355	0.6367
	0.6852	0.7006	0.6693	0.7528		0.6456	0.6825	0.6078	0.6931
	0.6367	0.6493	0.6238	0.6884		0.6000	0.7323	0.4658	0.6283
	0.6889	0.6920	0.6857	0.7593		0.6570	0.7170	0.5938	0.7062
	0.6847	0.7013	0.6687	0.7567		0.6513	0.7219	0.5836	0.7054
	0.6796	0.6985	0.6610	0.7505		0.6518	0.6990	0.6054	0.7008
	0.6968	0.7159	0.6775	0.7702		0.6681	0.7100	0.6256	0.7154
Average	0.6792	0.6933	0.6650	0.7444	Average	0.6457	0.7184	0.5718	0.6901
Standard deviation	0.0202	0.0195	0.0232	0.0280	Standard deviation	0.0229	0.0239	0.0652	0.0310