Table 6 Qualitive evaluation of the effect of the feature engineering method and clustering algorithm. We conducted Kruskal-Wallis test and effect size used in the table is partial \(\eta ^2\). Furthermore, we calculated Kendall’s W concordance index
From: DiviK: divisive intelligent K-means for hands-free unsupervised clustering in big biological data
Quality measure | Feature engineering | Clustering |
|---|---|---|
Partial \(\eta ^2\) | ||
 Rand index | 0.203 | 0.258 |
 Dice index | 0.161 | 0.293 |
 EXIMS | 0.262 | 0.095 |
 Overall quality d(0, 0, 0) | 0.141 | 0.345 |
 Overall quality d(1, 1, 1) | 0.122 | 0.309 |
Kendall’s W Concordance Index | ||
 Rand index | 0.328 | 0.325 |
 Dice index | 0.328 | 0.325 |
 EXIMS | 0.136 | 0.375 |
 Overall quality d(0, 0, 0) | 0.424 | 0.138 |
 Overall quality d(1, 1, 1) | 0.472 | 0.138 |