Slide 43
Slide 43 text
Confidential 43
bajji scoreʢ৴པείΞʣͷࢉग़ͷΞϧΰϦζϜ
bsi
= ∑
l
we
ij
+ ∑
m
˜
wim
+ ∑
n
˜
wni
˜
wb
ij
= ∑
k
wb
ijk
˜
we
ij
= λe
ij
⋅ ∑
k
we
ijk
wb
ijk
= μk
⋅ λb
ij
we
ijk
= min
(
1,
δe
ijk
30 )
λb
ijk
= exp (ln(CPy ⋅ 365 / δb
ijk
) ⋅ (1 − γij
) ⋅ Δγ) − 1
λe
ij
= exp (ln(CPy ⋅ 365 / δe
ij
) ⋅ (1 − γe) ⋅ Δγ) − 1
μk
=
{
ln(priceb
k
) (if 0 < priceb
k
AND bk
is sent)
priceb
k
(if 0 < priceb
k
AND bk
is received)
γij
=
{
σij
(1 / rsij
) (if CPd < δb
k
)
0 (otherwise)
δb
ijk
= ⌊
tcurrent − tb
ijk
86400
⌋ − CPd (if 0 < δb
k
)
1 (otherwise)
δe
ij
=
⌊
tcurrent − te
ij
86400
⌋ − CPd (if 0 < δe
ij
)
1 (otherwise)
δe
ijk
=
⌊
te
ijk
− te
ij(k−1)
86400
⌋ (if 1 < k)
⌊
tcurrent
− te
ijk
86400
⌋ (otherwise)
rsij
=
̂
εij
+ B ⋅ ̂
kij
δe
ij
σij
(x) = tanh (
x
x + 1 )
bsi
: bajji score of Ai
(Ai
: accounti
)
rsij
: relationship score between Ai
and Aj
B : bajji constant (B = 8.8)
CPy : bajji critical point year (CPy = B)
CPd : bajji critical point date (CPd = 10B)
Δγ : Aging correction (Δγ = B/1,000)
̂
uεi
: Number of encounters between Ai
and unique users
̂
ug
i
: Number of bajjis sent from Ai
to unique users
̂
uri
: Number of bajjis received by Ai
from unique users
̂
εi
: Number of encounters between Ai
and users
̂
gi
: Number of bajjis sent from Ai
to users
̂
ri
: Number of bajjis received by Ai
from users
˜
wij
: Sum of the value of the bajji sent from Ai
to Aj
wijk
: Value of the bajjik
sent from Ai
to Aj
priceb
k
: Price of the bajjik
tcurrent : Current datetime [sec]
̂
εij
: Number of encounter between Ai
and Aj
̂
kij
: Number of bajji send Ai
to Aj
tb
k
: Datetime of the bajjik
sent and received [sec]
te
ij
: Datetime of the last encounter between Ai
and Aj
[sec]
γe : Aging rate of encounter (γe = 0.0)
γij
: Aging rate between Ai
and Aj
δb
k
: Difference of days since the bajjik
sent and received between Ai
and Aj
δe
ij
: Difference of days since the last encountered between Ai
and Aj