Information Technology Reference
In-Depth Information
of drunken driving are severe, those who drink alcohol and drive are rare. However,
p
(
|
)
is a high value because When alcohol is drunk, he/she will get drunker and
driving a car will also become very dangerous.
This example shows the dependence about
B
of the anteroposterior correlation
between
A
and
B
. We call it a degree of dependence.
The following formula holds when event
A
and event
B
are independent:
A
B
p
(
A
,
B
)=
p
(
A
)
p
(
B
)
.
(3)
The following formula holds when event
A
and event
B
are not independent:
p
(
A
,
B
)=
p
(
A
|
B
)
p
(
B
)
.
(4)
. The ratio is a good
evaluation index for dependency. Therefore, the degree of dependence about
B
of
the anteroposterior correlation between
A
and
B
represents as follows a formula:
By two formulas, we focus on the ratio of
p
(
A
)
and
p
(
A
|
B
)
(
|
)
p
A
B
degreeO f Dependence
p
(
B
)
p
(
A
|
B
)
=
.
(5)
p
(
A
)
The higher evaluation index value presupposes that the degree of dependence
about
B
of the anteroposterior correlation between
A
and
B
is higher. This shows that
a possibility that event
B
has caused event
A
is higher. That is, the higher evaluation
index value presupposes better anteroposterior correlation.
Both Comparison of the conditional probability described in section 4.1 and the
evaluation index of the degree of dependence are used, and the system based on our
method extract appropriate anteroposterior correlation.
4.3
Application to Query Log
We apply our method to a query log. The nuisance variable
A
expresses the phe-
nomenon which inputs the query of a certain phenomenon
X
. The nuisance variable
B
shows that a certain phenomenon
Y
occurs. The
t
B
is the time when the phe-
nomenon
B
occurs.
In this situation, the probability
p
(
A
)
represents as follows a formula:
time
n
A
(
time
)
(
)=
)
,
p
A
time
N
A
(
(6)
time
where
N
(
time
)
is all the number of times of all the query of all the time,
n
A
(
time
)
is
the number of times of the query phenomenon
X
. The probability
p
(
A
|
B
)
represents
as following formula:
t
const
t
B
n
A
(
time
)
p
(
A
,
B
)
p
(
A
|
B
)=
=
)
,
(7)
)
t
const
t
B
p
(
B
)
p
(
B
N
A
(
time
Search WWH ::
Custom Search