Skip to main content

Table 3 Experiment results on 5 different models including 11 performance evaluation metrics

From: Application of multi-label classification models for the diagnosis of diabetic complications

Metric Traditional model The MLC models
BR LP CC ECC CLR
Example-based Metrics      
 Hamming loss 0.1864 0.2141 0.176 0.176 0.1763
 Accuracy 0.6816 0.6364 0.6948 0.702 0.6875
 F1_score 0.7661 0.721 0.7763 0.7855 0.7711
 Precision 0.8163 0.7706 0.8298 0.8649 0.8156
 Recall 0.78 0.7406 0.7861 0.7727 0.792
Label-based Metrics      
 F1_micro 0.789 0.7559 0.8000 0.8078 0.7968
 F1_macro 0.7593 0.708 0.7714 0.7773 0.7631
 Precision_micro 0.7709 0.7396 0.7757 0.7592 0.7863
 Precision_macro 0.7884 0.6822 0.7964 0.7689 0.8073
 Recall_micro 0.806 0.7731 0.8261 0.8631 0.8077
 Recall_macro 0.7476 0.7386 0.764 0.8009 0.7394
  1. The MLC models included a Label Power Set (LP), Classifier Chains (CC), Ensemble Classifier Chains (ECC), and Calibrated Label Ranking (CLR). The traditional model Binary Relevance (BR) is used as a comparison. The best performing method is in bold