Information Technology Reference
In-Depth Information
① based on similar users
correspond to user-based CF
probability of act
t
P
u
a
→act
t
?
② based on action
correspond to item-based CF
Fig. 6.
Approach of predicting missing activity
4.1
Prediction Based on Similar Users
Based on the following ideas, we calculate similarity between two users in emergency
situations.
-
It is highly probable that as same as user
u
a
, similar users also did before action
(
Did
(
a
before
)
) and after action (
Did
(
a
after
)
) of the candidate action
act
t
D
-
If users had the same goal (e.g. wanted to evacuate in Shinjuku), then they had same
action patterns (
SameTarget
(
a
t
,l
t
)
).
-
It is highly probable that user did the same actions if they were in the same location
(
SameLocation
(
l
)
).
Therefore, the similarity between user
u
j
and user
u
a
will be calculated as Equation 1.
S
(
u
j
,u
a
)=
βDid
(
{
a
before
,l
before
}
,
{
a
after
,l
after
}
)
+
γSameTarget
(
a
t
,l
t
)
+(1
(1)
−
β
−
γ
)
SameLocation
(
l
)
Where:
-
Parameters
β,γ
satisfy
0
≤
β,γ,β
+
γ
≤
1
. These parameters depend on each
particular problem.
-
If
u
j
did action
act
t
in location
l
,then
Did
(
act
t
,l
)=1
,otherwise
Did
(
act
t
,l
)=
0
.
-
If
u
j
and user
u
a
has the same goal (want to do action
a
t
in target location
l
t
), then
SameTarget
(
a
t
,l
t
)=1
,otherwise
SameTarget
(
a
t
,l
t
)=0
.
-
If
u
j
and user
u
a
were in the same location
l
at the time
t
,then
SameLocation
(
l
)=
1
,otherwise
SameLocation
(
l
)=0
.
4.2
Prediction Based on Probability of Action
In real-world, an action depends on location, time and its before-after actions. There-
fore, probability of
act
t
at the time
t
can be calculated as Equation 2.
P
(
act
t
)=
ρ
a
{
F
(
a
before
→
act
t
)+
F
(
act
t
→
a
after
)
}
(2)
+
ρ
t
F
(
act
t
,t
)+(1
−
ρ
a
−
ρ
t
)
F
(
act
t
,l
)
