Skip to main content

Advertisement

Table 2 Predictive performance of the seven different multi-label algorithms based on a variety of measures

From: Multi-label classification of music by emotion

  BR LP RAk EL 2BR CLR ML-k NN BP-MLL
Hamming Loss 0.1943 0.1964 0.1849 0.1953 0.1930 0.2616 0.2064
Accuracy 0.5185 0.5887 0.5876 0.5293 0.5271 0.3427 0.5626
Precision 0.6677 0.6840 0.7071 0.6895 0.6649 0.5184 0.6457
Recall 0.5938 0.7065 0.6962 0.6004 0.6142 0.3802 0.7234
F 1 0.6278 0.6945 0.7009 0.6411 0.6378 0.4379 0.6814
Subset acc. 0.2759 0.3511 0.3395 0.2839 0.2830 0.1315 0.2869
Micro prec. 0.7351 0.6760 0.7081 0.7280 0.7270 0.6366 0.6541
Micro rec. 0.5890 0.7101 0.6925 0.5958 0.6103 0.3803 0.7189
Micro F1 0.6526 0.6921 0.6993 0.6540 0.6622 0.4741 0.6840
Micro AUC 0.7465 0.8052 0.8241 0.7475 0.8529 0.7540 0.8474
Macro prec. 0.6877 0.6727 0.7059 0.6349 0.7036 0.4608 0.6535
Macro rec. 0.5707 0.7018 0.6765 0.5722 0.5933 0.3471 0.7060
Macro F1 0.6001 0.6782 0.6768 0.5881 0.6212 0.3716 0.6681
Macro AUC 0.7343 0.8161 0.8115 0.7317 0.8374 0.7185 0.8344
One-error 0.3038 0.3389 0.2593 0.2964 0.2512 0.3894 0.2946
Coverage 2.4378 1.9300 1.9983 2.4482 1.6914 2.2715 1.7664
Ranking loss 0.2776 0.1867 0.1902 0.2770 0.1456 0.2603 0.1635
Avg. precis. 0.7378 0.7632 0.7983 0.7392 0.8167 0.7104 0.7961