Information Technology Reference
In-Depth Information
(a) dataset13;
M
=10;
k
=3;
11754333
(b) dataset13;
M
=10;
k
=3;
11754333
(c) dataset15;
M
=18;
k
=2;
55 37 16 7 5 321
(d) dataset22;
M
=14;
k
=2;
45 26 12 8 5 3 22
(e) dataset34;
M
=20;
k
=3; 68 59
36251585322
(f) spiral;
M
=30;
k
=2;
116 72 28 14 5 3 2 1
Fig. 6.32 The clustering solutions for each dataset suggested by LEGClust. Each
label shows: the dataset name; the number of neighbors (
M
); the number of connec-
tions to join clusters (
k
); the number of clusters found at each step of the algorithm
(underlined is the step of the shown solution).
The DBScan algorithm failed in finding the reference clusters in all artifi-
cial datasets (except the one of Fig. 6.31a).
The results reported in [204] for all algorithms applied to the artificial datasets
lead to the conclusion that, as expected, the solutions obtained with the density-
based algorithms are worse than those obtained with any of the other algo-
rithms. The best results were obtained with LEGClust and Chameleon.
