Information Technology Reference
In-Depth Information
.
<λ
≤
.
(4) If 0
673
0
755, then
{
y
1
,
y
6
}
,
{
y
2
,
y
3
,
y
7
,
y
8
}
,
{
y
4
,
y
9
}
,
{
y
5
,
y
10
}
(5) If 0
.
755
<λ
≤
0
.
784, then
{
y
1
,
y
6
}
,
{
y
2
,
y
3
,
y
7
}
,
{
y
8
}
,
{
y
4
,
y
9
}
,
{
y
5
,
y
10
}
(6) If 0
.
784
<λ
≤
0
.
837, then
{
y
1
,
y
6
}
,
{
y
2
,
y
3
}
,
{
y
5
}
,
{
y
7
}
,
{
y
8
}
,
{
y
10
}
,
{
y
4
,
y
9
}
(7) If 0
.
837
<λ
≤
0
.
859, then
{
y
1
,
y
6
}
,
{
y
2
}
,
{
y
3
}
,
{
y
5
}
,
{
y
7
}
,
{
y
8
}
,
{
y
10
}
,
{
y
4
,
y
9
}
(8) If 0
.
859
<λ
≤
0
.
918, then
{
y
1
}
,
{
y
2
}
,
{
y
3
}
,
{
y
5
}
,
{
y
6
}
,
{
y
7
}
,
{
y
8
}
,
{
y
10
}
,
{
y
4
,
y
9
}
(9) If 0
.
918
<λ
≤
1, then
{
y
1
}
,
{
y
2
}
,
{
y
3
}
,
{
y
4
}
,
{
y
5
}
,
{
y
6
}
,
{
y
7
}
,
{
y
8
}
,
{
y
9
}
,
{
y
10
}
From the above numerical analysis, we know that Algorithm 2.1 only takes into
account the maximal and minimal deviation information, and ignores all the other
deviation information, more importantly, it cannot take into account any information
on attribute weights, and thus produces the loss of too much information, while
Algorithm 2.2 can not only avoid losing the given information, but also require less
computational effort and is more convenient in practical applications.
Now we further compare Algorithm 2.2 with Algorithm-FSC on the simulated
data set:
We first exploit Algorithm 2.2 on the simulated data set. In the experiment, we
set a series of
values ranging from 0.6 to 1.0, and compute the values of the SI
measure for each clustering result. The results can be found in Table
2.4
(Xu et al.
2008):
λ
Table 2.4
The results derived by Algorithm 2.2 with different
λ
levels on the simulated data set
λ
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
SI
0.437
0.437
0.437
0.437
0.437
0.437
0.437
0.437
0.995
K
3
3
3
3
3
3
3
3
900
Note
:(1)
K
is the number of clusters found by Algorithm 2.2.
(2) Since
C
2
7
C
2
6
, we get the equivalent associate matrix
C
2
6
after six iterations.
=
Search WWH ::
Custom Search