Information Technology Reference
In-Depth Information
Suppose
ň
s`1-s`k
and
ň
t`1-t`k
-elements and
total order, and that sim(.) is the similarity measurement function. If
are subset of
I
s1-sm
I
t1-tn
with
k
Γ
≈
Γ
∀
I
∈
Γ
,
∀
I
∈
Γ
,
(
I ,I
)
≥
δ
sim
,
written
as
,
s i
′
s -s k
′
1
′
t i
′
t
′
1
-t k
′
s
′
t
′
(1
≤ ≤
i k
)
s -s k
′
1
′
t
′
1
-t k
′
i
i
Ō
totally similar, where
Ō
then
I
s1-sm
and
I
t1-tn
are
k-
is the predefined
and
Ō
are, the
threshold value between 0 and 1. Obviously, the bigger
k
bigger degree of the similarity between
I
s1-sm
and
I
t1-tn
.
Suppose
C`
s`1-s`k
and
C`
t`1-t`k
are
k
-elements subsets of condition sequences
′
∀
C
∈
C
,
∀
C
∈
C
C
s1-sm
={
C
s1
,…,C
sm
}
and
C
t1-tn
={
C
t1
,…,C
tm
}.
If
,
s i
′
s -s k
′
1
′
t i
′
t
′
1
-t k
′
C
≈
C
(
C
,C
)
≥
δ
sim
, written as
, then
I
s1-sm
and
I
t1-tn
are
s i
′
t i
′
(1
≤ ≤
i
k
)
s -s k
′
1
′
t
′
1
-t k
′
conditional similar, where
Ō
is the predefined threshold value between 0
and 1.
Suppose
Ō
k-
I
u
,I
u+1
,…,I
cur
} is a fishing ground sequence of length
cur-u+1 from time u to time cur, and
I
u-cur
={
I
u-cur
, I
cur+1
}
represents the evolvement of fishing grounds from time u to the time cur+1,
where
I
u-cur+1
={
I
u
,I
u+1
,…,I
cur
,I
cur+1
}={
C
cur
+1
can be gathered accurately in the way that
the meteorological phenomena is analyzed, and
I
cur
+1
=(
C
cur
+1
,
G
cur
+1
),
G
cur
+1
is the fishing ground in
next week needed to be predicted. Suppose
ň
u'1-u'k
is a total order subset of
I
u-cur
+1
with elements
I
cur
and
I
cur
+1
. Our task is to find a history sequence
I
v-w
Γ
≈
Γ
with sub-sequence
ň
v'1-v'k
k-
δ totally similar to
ň
u'-u'k
.
u
′
1
-u k
′
v
′
1
-v k
′
′
′
C
≈
C
G
=
G
≈
G
implies
, which implies in turn
. Finally,
u
′
1
-u k
′
v
′
1
-v k
′
cur
+1
u k
′
v' k
we can fix the position of
G
v'k
.
As demonstrated by experiments, the condition for totally similar is very
strong. The notion of conditional similarity serves well. Suppose
G
cur
+1
by
C'
u'1-u'k
is a
total order subset of
C
u-cur
+1
with elements
C
cur
and
C
cur+1
. Our task is to find a
history sequence
C
v-w
with sub-sequence
C`
v'1-v'k
k-
δ conditionally similar to
′
′
C
≈
C
G
=
G
≈
G
C`
u'1-u'k.
.
implies
, then
G
cur
+1
=
G
cur
+
△
G
u
′
1
-u k
′
v
′
1
-v k
′
cur
+1
u k
′
v' k
and
△
G
G
v'k
-1
are the changes fishing grounds of this week takes to
form fishing grounds next week. This it is the very reason we introduce
several similarity measurements in previous section.
Due to the fact that it is difficult to set the value of, we adopt
=
G
v'k
-
- nearest
neighbor method to retrieve k nearest neighbors, based on which the central
fishing ground of next week can be calculated with great robustness and no
value of
k
Ō
.
