Skip to main content

Table 3. Comparison between the different clustering algorithms on the five datasets according to the NMI measure

From: Considerably Improving Clustering Algorithms Using UMAP Dimensionality Reduction Technique: A Comparative Study

  MNIST F-MNIST UMIST face Pen digits USPS
K-means 0.4774 0.5139 0.6647 0.6998 0.6266
UMAP + K-means 0.8494 0.6377 0.8663 0.8545 0.8602
Improvement score 0.3720 0.1238 0.2016 0.1547 0.2336
Agglomerative 0.6360 0.6080 0.6673 0.7965 0.7250
UMAP + Agglomerative 0.8463 0.6511 0.8764 0.8456 0.9000
Improvement score 0.2103 0.0431 0.2091 0.0491 0.1750
HDBSCAN 0.3674 0.2535 0.6933 0.5804 0.4442
UMAP + HDBSCAN 0.8315 0.6323 0.8427 0.8871 0.8923
Improvement score 0.4641 0.3788 0.1494 0.3067 0.4481
GMM 0.3882 0.5471 0.6160 0.5203 0.4232
UMAP+GMM 0.8654 0.6424 0.8648 0.8447 0.8231
Improvement score 0.4772 0.0953 0.2488 0.3244 0.3999
Back to paper page