A Decoupling and Bidirectional Resampling Method for Multilabel Classification of Imbalanced Data with Label Concurrence

Table 3

Macro-F values of different resampling algorithms.

Algorithm	Dataset	LP-ROS	REMEDIAL-HwR-HUS	REMEDIAL-HwR-ROS	ML-DBR
LP	Yeast	0.4328	0.4592	0.4629	0.4609
LP	Enron	0.3582	0.3863	0.3927	0.4151
LP	tmc-2007	0.5174	0.6138	0.6199	0.6264
LP	cal-500	0.3041	0.3028	0.3325	0.3406
LP	Corel-16k	0.0924	0.1057	0.1129	0.1377
LP	Corel-5k	0.1236	0.1429	0.1354	0.1438
LP	Mediamill	0.2248	0.2379	0.2406	0.2428
BR	Yeast	0.4320	0.4573	0.4512	0.4556
BR	Enron	0.3602	0.3667	0.4021	0.4134
BR	tmc-2007	0.5228	0.5854	0.6039	0.6194
BR	cal-500	0.3112	0.2935	0.3209	0.3461
BR	Corel-16k	0.0941	0.1124	0.1136	0.1461
BR	Corel-5k	0.1265	0.1479	0.1323	0.1468
BR	Mediamill	0.2392	0.2416	0.2473	0.2605
ML-kNN	Yeast	0.4292	0.4760	0.5374	0.5360
ML-kNN	Enron	0.4338	0.3851	0.4417	0.4624
ML-kNN	tmc-2007	0.4864	0.6133	0.6420	0.6515
ML-kNN	cal-500	0.3543	0.3691	0.3757	0.4204
ML-kNN	Corel-16k	0.1245	0.1263	0.1565	0.1829
ML-kNN	Corel-5k	0.2012	0.2205	0.2187	0.2274
ML-kNN	Mediamill	0.2572	0.2938	0.2856	0.3067