Research Article
Clustering by Detecting Density Peaks and Assigning Points by Similarity-First Search Based on Weighted K-Nearest Neighbors Graph
Table 3
The comparison of ACC, AMI, and ARI benchmarks for 6 clustering algorithms on synthetic datasets.
| Algorithm | AMI | ARI | ACC | EC/AC | Par | AMI | ARI | ACC | EC/AC | Par | Pathbased | Jain |
| DPC-SFSKNN | 0.926 | 0.910 | 0.925 | 3/3 | 6 | 1.000 | 1.000 | 1.000 | 2/2 | 7 | DPC | 0.521 | 0.463 | 0.742 | 3/3 | 2 | 0.609 | 0.713 | 0.853 | 2/2 | 3 | DBSCAN | 0.781 | 0.522 | 0.667 | — | 0.05/6 | 0.883 | 0.985 | 0.918 | — | 0.08/10 | AP | 0.679 | 0.475 | 0.783 | 3/3 | 10 | 0.681 | 0.812 | 0.882 | 2/2 | 40 | FKNN-DPC | 0.941 | 0.960 | 0.987 | 3/3 | 5 | 0.056 | 0.132 | 0.793 | — | 10 | K-means | 0.568 | 0.461 | 0.772 | — | 3 | 0.492 | 0.577 | 0.712 | — | 2 |
| | Aggregation | | | | | Flame | | | | |
| DPC-SFSKNN | 0.942 | 0.951 | 0.963 | 7/7 | 6 | 0.873 | 0.934 | 0.956 | 2/2 | 6 | DPC | 1.000 | 1.000 | 1.000 | 7/7 | 4 | 1.000 | 1.000 | 1.000 | 2/2 | 5 | DBSCAN | 0.969 | 0.982 | 0.988 | — | 0.05/8 | 0.867 | 0.936 | 0.981 | - | 0.09/8 | AP | 0.795 | 0.753 | 0.841 | 7/7 | 7.7 | 0.452 | 0.534 | 0.876 | 3/3 | 35 | FKNN-DPC | 0.995 | 0.997 | 0.999 | 3/3 | 8 | 1.000 | 1.000 | 1.000 | 2/2 | 5 | K-means | 0.784 | 0.717 | 0.786 | — | 7 | 0.418 | 0.465 | 0.828 | — | 2 |
| | DIM512 | | | | | DIM1024 | | | | |
| DPC-SFSKNN | 1.000 | 1.000 | 1.000 | 16/16 | 8 | 1.000 | 1.000 | 1.000 | 16/16 | 9 | DPC | 1.000 | 1.000 | 1.000 | 16/16 | 2 | 1.000 | 1.000 | 1.000 | 16/16 | 0.01 | DBSCAN | 1.000 | 1.000 | 1.000 | — | 0.3/7 | 1.000 | 1.000 | 1.000 | — | 10/8 | AP | 1.000 | 1.000 | 1.000 | 16/16 | 20 | 1.000 | 1.000 | 1.000 | 16/16 | 30 | FKNN-DPC | 1.000 | 1.000 | 1.000 | 16/16 | 8 | 1.000 | 1.000 | 1.000 | 16/16 | 10 | K-means | 0.895 | 0.811 | 0.850 | — | 1 | 0.868 | 0.752 | 0.796 | — | 16 |
|
|