Skip to main content

Table 5 High-level summary of feature engineering characteristics

From: DiviK: divisive intelligent K-means for hands-free unsupervised clustering in big biological data

 

UMAP

Xception

Knee PCA

EXIMS PCA

None

Rand index ranks

48.5

52.5

49

25

35

Dice index ranks

48

51

49

28

34

EXIMS score ranks

51.5

32.5

29.5

60

36.5

Scalability ranks

38

58

38

38

38

Total rank

186

194

165.5

151

143.5

  1. Contains sum of ranks for each criterion (lower is better). Top rank of a quality index per clustering method is presented in bold italics