Databases Reference
In-Depth Information
Memory
10
2
Mushroom: Our algorithm
Mushroom: Closet+
COG
Chess
10
1
10
0
0
0.2
0.4
0.6
0.8
1
minimum weighted density
Fig. 6.
Memory comparison
Mushroom
10
Our algorithm
CLOSET+
8
6
4
2
0
0
0.2
0.4
0.6
0.8
1
minimum weighted density
Fig. 7.
Mushroom time comparison
We also want to show through experiments that using weighted density
can find more clusters in less dense subspaces. So we compared the results
from density pruning with the results from weighted density pruning. For fair
comparison, we only compare when the minimum density threshold and the
minimum weighted density threshold are equally selective, that is, there are
equal number of clusters that satisfy each of the constraint. Figure 10 shows
the percentage of the clusters being found after each attribute id on COG
