Table 3 Metrics adopted for clustering evaluation
Name | Acronym | Computed as | Codomain | Better if |
|---|---|---|---|---|
Rand’s Index | RI | \(\frac {TP+TN}{TP+FP+FN+TN}\) | [0,1] | ↑ |
Adjusted Rand’s Index | ARI | \(\frac {RI-E[RI]}{\max (RI)-E[RI]}\) | [-1,1] | ↑ |
Misclassification Rate | MR | \(\frac {FP+FN}{TP+FP+FN+TN}\) | [0,1] | ↓ |
F-Measure | F1 | \(\frac {2 Precision \cdot Recall}{Precision + Recall}\) | [0,1] | ↑ |
Fowlkes-Mallows Index | FMI | \(\frac {TP}{\sqrt {(TP + FP)\cdot (TP+ FN)}}\) | [0,1] | ↑ |
Cluster Accuracy | CA | \(\frac {1}{n} \sum _{i=1}^{c} \max ({CP}_{i}|{GT}_{i})\) | (0,1] | ↑ |
Normalized Mutual Information | NMI | \(\frac {MI}{\sqrt {H(CP) H(GT)}}\) | [0,1] | ↑ |
Adjusted Mutual Information | AMI | \(\frac {MI-E[MI]}{\sqrt {H(CP) H(GT)}-E[MI]}\) | [-1,1] | ↑ |