From: DiviK: divisive intelligent K-means for hands-free unsupervised clustering in big biological data
Global feature engineering method | Adjusted Rand index | Dice index | Relative EXIMS score | Overall quality d(0, 0, 0) | Overall quality d(1, 1, 1) |
---|---|---|---|---|---|
UMAP | 0.364 (0.288) | 0.511 (0.371) | 0.742 (0.180) | 1.052 (0.190) | 0.881 (0.407) |
Xception | 0.346 (0.271) | 0.487 (0.361) | 0.886 (0.138) | 1.139 (0.124) | 0.868 (0.397) |
Knee PCA | 0.447 (0.123) | 0.538 (0.359) | 0.871 (0.144) | 1.167 (0.105) | 0.766 (0.315) |
EXIMS PCA | 0.585 (0.094) | 0.725 (0.151) | 0.703(0.090) | 1.176 (0.100) | 0.596 (0.119) |
None | 0.508 (0.116) | 0.703 (0.134) | 0.846 (0.168) | 1.227 (0.091) | 0.620 (0.141) |