Database Reference
In-Depth Information
p
=
0
.
5
p
=
0
.
7
RID 0
.
5-rank # of Days RID 0
.
7-rank # of Days
R
1
2
435
.
8
R
1
2
435
.
8
R
2
2
341
.
7
R
2
2
341
.
7
R
3
3
335
.
7
R
3
3
335
.
7
R
4
4
323
.
9
R
5
5
284
.
7
R
5
4
284
.
7
R
6
6
266
.
8
R
6
5
266
.
8
R
9
7
233
.
6
R
9
7
233
.
6
R
11
9
232
.
6
R
10
8
233
.
3
R
10
10
233
.
3
R
11
8
232
.
6
R
18
13
229
.
3
R
14
10
231
.
1
R
23
15
227
.
2
Table 5.5
Results of Top-
(
p
,
l
)
queries on the IIP Iceberg Sighting Database (
l
=10).
ship probability values and the top-10 probability values of some tuples including
the ones returned by the PT-
k
, U-Top
k
, and U-
K
Ranks queries.
All tuples with a top-10 probability of at least 0
5 are returned by the PT-
k
query.
The top-10 probability of
R
14 is higher than
R
7, but
R
7 is included in the answer
of the U-Top
k
query and
R
14 is missing. Moreover, the presence probability of
the top-10 list returned by the U-Top
k
query is quite low. Although it is the most
probable top-10 tuple list, the low presence probability limits its usefulness and
interestingness.
R
10 and
R
14, whose top-10 probability values are high, are missing in the re-
sults of the U-
K
Ranks query, since none of them is the most probable at any rank.
Nevertheless,
R
18 is returned by the U-
K
Ranks query at the 10-th position, though
its top-10 probability is much lower than
R
10 and
R
14. Moreover,
R
9 and
R
11 each
occupies two positions in the answer of the U-
K
Ranks query.
The results clearly show that the
PT-k query
captures some important tuples
missed by the
U-TopK query
and the
U-KRanks query
.
.
5.6.1.2 Answering Top-(
k
,
l
) Queries, PT-
k
Queries and Top-(
p
,
l
) Queries
Moreover, We conduct top-(
k
,
l
) queries, PT-
k
queries and top-(
p
,
l
) queries on the
database.
Table 5.4 shows the results of a top-
query. To un-
derstand the answer, the top-5 probabilities, the top-20 probabilities, and the number
of days drifted are also included in the table. Some records returned by the top-
(
(
5
,
10
)
query and a top-
(
20
,
10
)
query, such as
R
4,
R
7,
R
10 and
R
8.
Moreover,
R
11,
R
18,
R
23 and
R
33 are returned by the top-
5
,
10
)
are not in the results of the top-
(
20
,
10
)
(
20
,
10
)
query but are not
in the results of the top-
query. When
k
becomes larger, the records ranked
relatively lower but with larger membership probabilities may become the answers.
It is interesting to vary value
k
and compare the difference among the query results.
The results to a top-
(
5
,
10
)
(
.
,
)
(
.
,
)
0
5
10
query and a top-
0
7
10
query are listed in Ta-
ble 5.5. The 0
.
5-ranks, 0
.
7-ranks and the number of days drifted are also included.