47 /342
臺大管理論叢
第
27
卷第
2S
期
47
Table 8 The Clusters: (a) w is Set as 0.1; (b) w is Set as 0.9
Medoid
Cluster members
[
A
4
:[361, 367]]
[
A
4
:[362, 365]] (1901.214)
[
A
4
:[361, 367]] (2182.797)
[
A
4
:[368, 374]] (3613.998)
[
A
2
:[335, 337],
A
4
:[368, 374]]
(1682.998)
[
A
5
:[192, 197]]
[
A
2
:[335, 337]] (2074.049)
[
A
5
:[190, 194]] (3692.028)
[
A
5
:[195, 197]] (1576.016)
[
A
5
:[188, 193]] (3365.006)
[
A
5
:[192, 197]] (2077.285)
[
A
3
:[33, 35],
A
4
:[362, 365],
A
5
:[190, 194]]
[
A
3
:[33, 35],
A
4
:[362, 365],
A
5
:[190, 194]](1614.752)
[
A
3
:[30, 34],
A
4
:[361, 367],
A
5
:[195, 197]](1575.002)
[
A
3
:[33, 35],
A
4
:[368, 374],
A
5
:[190, 194]] (3100.002)
[
A
3
:[34, 37]]
[
A
3
:[35, 37]] (1775.109)
[
A
3
:[34, 37]] (2662.734)
[
A
3
:[33, 35]] (3018.757)
[
A
1
:[427, 435],
A
3
:[34, 37]]
(1590.300)
[
A
3
:[30, 33]]
[
A
1
:[427, 433]] (1707.000)
[
A
1
:[427, 435]] (1876.000)
[
A
3
:[30, 33]] (1736.733)
[
A
3
:[30, 34]] (2315.562)
(a)
Medoid Cluster members
[
A
3
:[35, 37]]
[
A
3
:[35, 37]] (1775.109)
[
A
3
:[30, 33]] (1736.733)
[
A
4
:[361, 367]]
[
A
2
:[335, 337]] (2074.049)
[
A
3
:[30, 34]] (2315.562)
[
A
4
:[362, 365]] (1901.214)
[
A
4
:[361, 367]] (2182.797)
[
A
5
:[192, 197]] (2077.285)
[
A
3
:[30, 34],
A
5
:[361, 367],
A
5
:[195, 197]]
[
A
5
:[195, 197]] (1576.016)
[
A
1
:[427, 435],
A
3
:[34, 37]](1590.300)
[
A
3
:[33, 35],
A
4
:[362, 365],
A
5
:[190, 194]](1614.752)
[
A
3
:[30, 34],
A
4
:[361, 367],
A
5
:[195, 197]](1575.002)
[
A
1
:[427, 433]]
[
A
1
:[427, 433]] (1707.000)
[
A
1
:[427, 435]] (1876.000)
[
A
2
:[335, 337],
A
4
:[368, 374]]
(1682.998)
[
A
3
:[33, 35],
A
5
:[368, 374],
A
5
:[190, 194]]
[
A
3
:[34, 37]] (2662.734)
[
A
3
:[33, 35]] (3018.757)
[
A
4
:[368, 374]] (3613.998)
[
A
5
:[190, 194]] (3692.028)
[
A
5
:[188, 193]] ( 3365.006)
[
A
3
:[33, 35],
A
5
:[368, 374],
A
5
:[190, 194]](3100.002)
(b)
If
w
is set as a high value, e.g., 0.9, each of the resulted clusters has a tight bound on the
expected supports of the contained FU2Ps, but the FU2Ps in a cluster have significantly
different appearances. The clusters derived by setting
w
,
ξ
, and
δ
as 0.9, 0.25, and 1.1,
respectively, are shown in Table 8(b).
Both settings presented in Table 8 provide informative summaries; however, the two
summaries have different properties. On one hand, using a low value for
w
often results in
more attributes appearing in a cluster because the appearances of the FU2Ps in the cluster
present more variations. On the other hand, using a high value for
w
gives a more
informative estimation for the expected support of a FU2P. The users may be interested in
the estimation for the expected support of a FU2P.
Figure 3 presents the formal algorithm of the SFC algorithm.