Database Reference
In-Depth Information
k
=
5
k
=
20
RID Top-5 prob. # of Days Drifted RID Top-20 Prob. # of Days Drifted
R
1
0
.
8
435
.
8
R
1
0
.
8
435
.
8
R
2
0
.
8
341
.
7
R
2
0
.
8
341
.
7
R
3
0
.
8
335
.
7
R
3
0
.
8
335
.
7
R
5
0
.
8
284
.
7
R
5
0
.
8
284
.
7
R
6
0
.
61
266
.
8
R
6
0
.
8
266
.
8
R
4
0
.
6
323
.
9
R
9
0
.
8
233
.
6
R
9
0
.
22
233
.
6
R
11
0
.
8
232
.
6
R
7
0
.
17
259
.
5
R
18
0
.
8
229
.
3
R
10
0
.
09
233
.
3
R
23
0
.
79
227
.
2
R
8
0
.
05
240
.
4
R
33
0
.
75
222
.
2
Table 5.4
Results of top-
(
k
,
l
)
queries on the IIP Iceberg Sighting Database (
l
=10).
determining the status of icebergs. It is interesting to find the icebergs drifting for a
long period.
However, each sighting record in the database is associated with a confidence
level according to the source of sighting, including: R/V (radar and visual), VIS
(visual only), RAD(radar only), SAT-L(low earth orbit satellite), SAT-M (medium
earth orbit satellite) and SAT-H (high earth orbit satellite). In order to quantify the
confidence, we assign confidence values 0
.
.
.
.
.
.
8, 0
7, 0
6, 0
5, 0
4 and 0
3 to the above
six confidence levels, respectively.
Moreover, generation rules are defined in the following way. For the sightings
with the same time stamp, if the sighting locations are very close - differences in
latitude and longitude are both smaller than 0
02 miles), they are con-
sidered referring to the same iceberg, and only one of the sightings is correct. All
tuples involved in such a sighting form a multi-tuple rule. For a rule
R
:
t
r
1
⊕···⊕
.
01 (i.e.,0
.
t
r
m
,
Pr
is set to the maximum confidence among the membership probability values
of tuples in the rule. Then, the membership probability of a tuple is adjusted to
Pr
(
R
)
t
r
l
)
∑
1
≤
i
≤
m
con f
con f
(
(
t
r
l
)=
Pr
(
R
)(
1
≤
l
≤
m
)
, where
con f
(
t
r
l
)
is the confidence of
t
r
l
.
(
t
r
i
)
After the above preprocessing, the database contains 4
231 tuples and 825 multi-
tuple rules. The number of tuples involved in a rule varies from 2 to 10. We name
the tuples in the number of drifted days descending order. For example, tuple
R
1 has
the largest value and
R
2 has the second largest value on the attribute.
,
5.6.1.1 Comparing PT-
k
queries, U-Top
k
queries and U-
K
Ranks Queries
We apply a
PT-k query
,a
U-TopK query
and a
U-KRanks query
on the database by
setting
k
5. The ranking order is the number of drifted days descend-
ing order. The PT-
k
query returns a set of 10 records
=
10 and
p
=
0
.
{
R
1
,
R
2
,
R
3
,
R
4
,
R
5
,
R
6
,
R
9
,
R
10
,
R
11
,
R
14
}
. The U-Top
k
query returns a vector
R
1
,
R
2
,
R
3
,
R
4
,
R
5
,
R
6
,
R
7
,
R
9
,
R
10
0299. The U-
K
Ranks query returns 10 tuples shown in
Table 5.3. The probability values of the tuples at the corresponding ranks are also
shown in the table. To understand the answers, in Table 5.2 we also list the member-
,
R
11
with probability 0
.
