I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
,
p
p
.
117
~
123
I
SS
N:
2252
-
8814
117
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
AAS
M
a
chine Lea
rnin
g
in
Big
Data
L
ido
ng
Wa
ng
*,
G
ua
ng
h
ui
Wa
ng
**,
Chery
l A
nn
Alex
a
nd
er
***
*
D
e
p
a
rt
m
e
n
t
o
f
En
g
in
e
e
rin
g
T
e
c
h
n
o
lo
g
y
,
M
issis
sip
p
i
V
a
ll
e
y
S
tate
Un
iv
e
rsity
,
US
A
*
*
S
t
a
te Ke
y
L
a
b
o
ra
to
ry
o
f
S
e
v
e
r
e
W
e
a
th
e
r,
Ch
in
e
se
A
c
a
d
e
m
y
o
f
M
e
teo
ro
l
o
g
ica
l
S
c
ien
c
e
s,
Ch
in
a
*
*
*
T
e
c
h
n
o
lo
g
y
a
n
d
He
a
lt
h
c
a
re
S
o
lu
ti
o
n
s
,
USA
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
28
,
2
0
1
5
R
ev
i
s
ed
No
v
7
,
2
0
1
5
A
cc
ep
ted
No
v
18
,
2
0
1
5
M
a
c
h
in
e
lea
rn
i
n
g
is
a
n
a
rti
f
icia
l
in
telli
g
e
n
c
e
m
e
th
o
d
o
f
d
isc
o
v
e
rin
g
k
n
o
w
led
g
e
f
o
r
m
a
k
in
g
in
telli
g
e
n
t
d
e
c
isio
n
s.
Big
Da
ta
h
a
s
g
re
a
t
i
m
p
a
c
ts
o
n
sc
ien
ti
f
ic
d
isc
o
v
e
rie
s
a
n
d
v
a
lu
e
c
re
a
ti
o
n
.
T
h
is
p
a
p
e
r
in
tro
d
u
c
e
s
m
e
th
o
d
s
in
m
a
c
h
in
e
lea
rn
in
g
,
m
a
in
tec
h
n
o
l
o
g
ies
i
n
Big
Da
ta,
a
n
d
so
m
e
a
p
p
l
ica
ti
o
n
s
o
f
m
a
c
h
in
e
lea
rn
in
g
in
Big
Da
ta
. C
h
a
ll
e
n
g
e
s o
f
m
a
c
h
in
e
lea
rn
in
g
a
p
p
l
ica
ti
o
n
s i
n
Big
Da
ta
a
re
d
isc
u
ss
e
d
.
S
o
m
e
n
e
w
m
e
th
o
d
s
a
n
d
tec
h
n
o
l
o
g
y
p
ro
g
re
ss
o
f
m
a
c
h
in
e
lea
rn
in
g
in
Big
Da
ta are
a
lso
p
re
se
n
ted
.
K
ey
w
o
r
d
:
B
ig
d
ata
B
ig
d
ata
an
al
y
tics
I
n
f
o
r
m
a
tio
n
tec
h
n
o
lo
g
y
Ma
ch
i
n
e
lear
n
i
n
g
Net
w
o
r
k
s
S
tr
ea
m
p
r
o
ce
s
s
i
n
g
Co
p
y
rig
h
t
©
201
5
In
s
t
it
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
L
id
o
n
g
W
an
g
,
Dep
ar
t
m
en
t o
f
E
n
g
i
n
ee
r
in
g
T
ec
h
n
o
lo
g
y
,
Miss
i
s
s
ip
p
i V
alle
y
State
U
n
i
v
er
s
it
y
,
US
A
.
E
m
ail:
l
w
a
n
g
2
2
@
s
t
u
d
en
ts
.
t
n
t
ec
h
.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
Ma
ch
i
n
e
lear
n
in
g
is
an
i
m
p
o
r
t
an
t
ar
ea
o
f
ar
ti
f
icia
l
i
n
telli
g
e
n
ce
.
T
h
e
o
b
j
ec
tiv
e
o
f
m
ac
h
in
e
l
ea
r
n
in
g
i
s
to
d
is
co
v
er
k
n
o
w
led
g
e
an
d
m
ak
e
in
telli
g
en
t
d
ec
is
io
n
s
.
Ma
ch
in
e
lear
n
i
n
g
a
lg
o
r
it
h
m
s
ca
n
b
e
ca
teg
o
r
ized
in
to
s
u
p
er
v
i
s
ed
,
u
n
s
u
p
er
v
i
s
ed
,
an
d
s
e
m
i
-
s
u
p
er
v
is
ed
.
W
h
e
n
b
i
g
d
ata
i
s
co
n
ce
r
n
ed
,
it
is
n
e
ce
s
s
ar
y
to
s
ca
le
u
p
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
it
h
m
s
[
1]
,
[2
]
.
A
n
o
t
h
er
ca
teg
o
r
izatio
n
o
f
m
ac
h
i
n
e
lear
n
i
n
g
ac
co
r
d
in
g
to
th
e
o
u
t
p
u
t
o
f
a
m
ac
h
in
e
lear
n
i
n
g
s
y
s
te
m
i
n
cl
u
d
es
clas
s
i
f
icatio
n
,
r
e
g
r
ess
io
n
,
clu
s
ter
i
n
g
,
a
n
d
d
en
s
i
t
y
e
s
ti
m
atio
n
,
etc.
Ma
ch
i
n
e
lear
n
in
g
ap
p
r
o
ac
h
es
i
n
clu
d
e
d
ec
is
io
n
tr
ee
lear
n
i
n
g
,
ass
o
c
iatio
n
r
u
le
lear
n
i
n
g
,
ar
ti
f
icial
n
eu
r
al
n
et
w
o
r
k
s
,
s
u
p
p
o
r
t v
ec
to
r
m
ac
h
in
e
s
(
SV
M)
,
clu
s
ter
in
g
,
B
a
y
e
s
ian
n
et
wo
r
k
s
,
an
d
g
e
n
etic
al
g
o
r
ith
m
s
,
e
tc.
,
[
3
]
.
E
x
am
p
le
s
o
f
s
u
p
er
v
is
ed
l
ea
r
n
in
g
alg
o
r
it
h
m
s
i
n
clu
d
e
N
aïv
e
B
a
y
es,
b
o
o
s
tin
g
alg
o
r
it
h
m
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
es
(
S
VM
)
,
an
d
m
ax
i
m
u
m
e
n
tr
o
p
y
m
e
th
o
d
(
Ma
x
E
NT
)
,
etc.
Un
s
u
p
er
v
i
s
ed
lear
n
i
n
g
tak
e
s
u
n
lab
elled
d
ata
a
n
d
class
i
f
ie
s
b
y
co
m
p
ar
i
n
g
t
h
e
f
ea
t
u
r
es
o
f
d
ata.
E
x
a
m
p
les
o
f
u
n
s
u
p
er
v
is
ed
alg
o
r
ith
m
s
ar
e
clu
s
ter
i
n
g
(
k
-
m
ea
n
s
,
d
en
s
it
y
-
b
ased
,
an
d
h
ier
ar
ch
ical,
etc.
)
,
s
elf
-
o
r
g
a
n
izi
n
g
m
ap
s
(
SO
M)
,
an
d
ad
ap
tiv
e
r
eso
n
an
ce
t
h
eo
r
y
(
A
R
T
)
[
4
]
.
Ma
ch
i
n
e
lear
n
i
n
g
h
a
s
b
ee
n
u
s
ed
in
b
ig
d
ata.
B
ig
d
ata
is
a
m
as
s
iv
e
v
o
lu
m
e
o
f
b
o
th
s
tr
u
ctu
r
ed
an
d
u
n
s
tr
u
ct
u
r
ed
d
ata
th
a
t
is
s
o
lar
g
e
t
h
at
it
i
s
d
i
f
f
ic
u
lt
to
p
r
o
ce
s
s
u
s
in
g
tr
ad
itio
n
a
l
d
ata
b
ase
an
d
s
o
f
t
w
ar
e
tech
n
iq
u
es.
B
i
g
d
ata
tec
h
n
o
l
o
g
ies
h
a
v
e
g
r
ea
t
i
m
p
ac
t
s
o
n
s
cien
t
if
ic
d
is
co
v
er
ie
s
a
n
d
v
al
u
e
cr
ea
tio
n
[
5
]
-
[
7
]
.
Ma
s
s
i
v
e
p
ar
allel
-
p
r
o
ce
s
s
in
g
(
MP
P),
d
is
tr
ib
u
ted
f
ile
s
y
s
te
m
s
,
a
n
d
clo
u
d
co
m
p
u
ti
n
g
,
e
tc.
ar
e
s
u
p
p
o
r
tin
g
tech
n
o
lo
g
ies
o
f
B
i
g
Data
[
8
]
.
B
esid
es
g
e
n
er
al
clo
u
d
i
n
f
r
a
s
tr
u
ctu
r
e
s
er
v
ice
s
,
tec
h
n
o
lo
g
ie
s
s
u
ch
as
Had
o
o
p
,
Data
b
ase
s
/Ser
v
er
s
SQ
L
,
No
S
QL
,
an
d
MP
P
d
atab
ases
,
etc.
a
r
e
also
u
s
ed
to
s
u
p
p
o
r
t B
ig
Data
[
9
]
.
T
h
is
p
ap
er
in
tr
o
d
u
ce
s
m
ac
h
i
n
e
lear
n
i
n
g
,
its
ap
p
licatio
n
s
in
B
ig
Data
,
a
n
d
th
e
c
h
all
en
g
e
s
an
d
tech
n
o
lo
g
y
p
r
o
g
r
es
s
o
f
m
ac
h
i
n
e
lear
n
i
n
g
in
B
i
g
Data
.
T
h
e
o
r
g
an
izatio
n
o
f
t
h
is
p
ap
er
is
as
f
o
llo
w
s
:
t
h
e
n
ex
t
s
ec
tio
n
i
n
tr
o
d
u
ce
s
m
eth
o
d
s
o
f
m
ac
h
i
n
e
lear
n
i
n
g
an
d
b
i
g
d
ata
;
Sectio
n
3
in
tr
o
d
u
ce
s
m
ac
h
i
n
e
lear
n
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
1
7
–
1
2
3
118
ap
p
licatio
n
s
i
n
b
ig
d
ata
;
Secti
o
n
4
d
is
cu
s
s
es
ch
al
len
g
es
o
f
m
ac
h
in
e
lear
n
i
n
g
ap
p
licatio
n
s
in
b
i
g
d
ata
;
Sectio
n
5
p
r
esen
ts
tec
h
n
o
lo
g
y
p
r
o
g
r
ess
o
f
m
ac
h
i
n
e
lear
n
in
g
ap
p
licatio
n
s
in
b
ig
d
ata
;
a
n
d
t
h
e
f
i
n
al
s
ec
tio
n
i
s
co
n
clu
s
io
n
s
.
2.
M
E
T
H
O
DS O
F
M
ACH
I
NE
L
E
A
RNIN
G
AN
D
B
I
G
DA
T
A
Su
p
er
v
i
s
ed
lear
n
i
n
g
ca
n
b
e
d
iv
id
ed
in
to
cla
s
s
i
f
icatio
n
an
d
r
eg
r
ess
io
n
.
W
h
en
th
e
c
lass
attr
ib
u
te
i
s
d
is
cr
ete,
it
is
ca
lled
class
i
f
ic
atio
n
;
w
h
e
n
t
h
e
clas
s
attr
ib
u
te
is
co
n
ti
n
u
o
u
s
,
it
i
s
r
eg
r
es
s
io
n
.
Dec
is
io
n
tr
ee
lear
n
in
g
,
n
aiv
e
B
a
y
es
cla
s
s
i
f
ier
,
k
-
n
ea
r
est
n
ei
g
h
b
o
r
(
k
N
N)
class
i
f
ier
,
an
d
cla
s
s
i
f
icat
io
n
w
it
h
n
et
w
o
r
k
in
f
o
r
m
atio
n
ar
e
class
if
ica
tio
n
m
et
h
o
d
s
.
L
in
ea
r
r
eg
r
e
s
s
io
n
an
d
lo
g
i
s
tic
r
eg
r
e
s
s
io
n
ar
e
r
eg
r
ess
io
n
m
et
h
o
d
s
.
Un
s
u
p
er
v
i
s
ed
lear
n
i
n
g
i
s
th
e
u
n
s
u
p
er
v
is
ed
d
iv
is
io
n
o
f
in
s
ta
n
ce
s
in
to
g
r
o
u
p
s
o
f
s
i
m
ilar
o
b
j
e
cts [
1
0
]
.
C
lu
s
ter
i
n
g
ca
n
b
e
g
r
o
u
p
ed
i
n
to
th
r
ee
ca
te
g
o
r
ies.
T
h
e
y
a
r
e
s
u
p
er
v
i
s
ed
,
u
n
s
u
p
er
v
is
ed
,
an
d
s
e
m
i
-
s
u
p
er
v
i
s
ed
[
1
1
]
:
1.
Su
p
er
v
i
s
ed
clu
s
ter
in
g
:
It
id
en
tif
ie
s
cl
u
s
ter
s
th
at
h
a
v
e
h
i
g
h
p
r
o
b
a
b
ilit
y
d
en
s
ities
w
it
h
r
es
p
ec
t
to
in
d
iv
id
u
al
cla
s
s
e
s
(
class
‐
u
n
if
o
r
m
cl
u
s
ter
s
)
.
I
t
i
s
u
s
ed
w
h
en
th
er
e
is
a
tar
g
et
v
ar
iab
le
a
n
d
a
tr
ain
i
n
g
s
et
t
h
a
t
in
cl
u
d
es th
e
v
ar
iab
les to
clu
s
te
r
.
2.
Un
s
u
p
er
v
i
s
ed
clu
s
ter
i
n
g
:
It
m
ax
i
m
izes
t
h
e
in
tr
ac
l
u
s
ter
s
i
m
i
lar
it
y
an
d
m
i
n
i
m
ize
s
th
e
in
ter
clu
s
ter
s
i
m
ilar
it
y
w
h
e
n
a
s
i
m
ilar
it
y
/
d
is
s
i
m
ilar
it
y
m
ea
s
u
r
e
is
g
i
v
e
n
.
I
t
u
s
es
a
s
p
ec
if
ic
o
b
j
ec
tiv
e
f
u
n
ctio
n
(
e.
g
.
,
a
f
u
n
ctio
n
t
h
at
m
in
i
m
izes t
h
e
i
n
tr
ac
lass
d
is
ta
n
ce
s
to
f
i
n
d
ti
g
h
t
clu
s
ter
s
)
.
K
‐
m
ea
n
s
a
n
d
h
ier
ar
ch
ical
cl
u
s
ter
i
n
g
ar
e
th
e
m
o
s
t
w
id
el
y
u
s
ed
u
n
s
u
p
er
v
is
ed
cl
u
s
ter
i
n
g
tec
h
n
iq
u
e
s
in
s
eg
m
e
n
tatio
n
.
3.
Se
m
i
-
s
u
p
er
v
i
s
ed
clu
s
ter
i
n
g
:
I
n
ad
d
itio
n
to
t
h
e
s
i
m
ilar
it
y
m
ea
s
u
r
e,
s
e
m
i
-
s
u
p
er
v
i
s
ed
clu
s
ter
in
g
u
tili
ze
s
o
th
er
g
u
id
in
g
/ad
j
u
s
t
in
g
d
o
m
a
in
i
n
f
o
r
m
atio
n
to
i
m
p
r
o
v
e
th
e
cl
u
s
ter
i
n
g
r
esu
lt
s
.
T
h
is
d
o
m
ai
n
in
f
o
r
m
atio
n
ca
n
b
e
p
air
w
is
e
co
n
s
tr
ain
t
s
b
et
w
ee
n
t
h
e
o
b
s
er
v
atio
n
s
o
r
tar
g
et
v
ar
iab
le
s
f
o
r
s
o
m
e
o
f
t
h
e
o
b
s
er
v
atio
n
s
.
Dec
is
io
n
tr
ee
s
cla
s
s
i
f
y
e
x
a
m
p
les
b
ased
o
n
th
eir
f
ea
t
u
r
e
v
alu
e
s
.
Dec
is
io
n
tr
ee
s
ar
e
co
n
s
tr
u
cte
d
r
ec
u
r
s
iv
e
l
y
f
r
o
m
tr
ain
i
n
g
d
ata
u
s
in
g
a
to
p
-
d
o
w
n
g
r
ee
d
y
ap
p
r
o
ac
h
in
w
h
ic
h
f
ea
t
u
r
es
ar
e
s
eq
u
en
tial
l
y
s
elec
ted
[
1
0
]
.
Dec
is
io
n
tr
ee
class
i
f
ier
s
o
r
g
an
ize
th
e
tr
ai
n
i
n
g
d
ata
in
to
a
tr
ee
-
s
tr
u
c
tu
r
e
p
la
n
.
Dec
is
io
n
tr
ee
s
ar
e
co
n
s
tr
u
cted
b
y
s
tati
n
g
w
it
h
t
h
e
r
o
o
t
n
o
d
e
h
a
v
in
g
th
e
w
h
o
le
d
ata
s
et,
iter
ati
v
el
y
c
h
o
o
s
in
g
s
p
litt
in
g
cr
iter
ia
a
n
d
ex
p
an
d
in
g
lea
f
n
o
d
es
w
it
h
p
ar
titi
o
n
ed
d
ata
s
u
b
s
ets
ac
co
r
d
in
g
to
th
e
s
p
li
tti
n
g
cr
iter
ia.
S
p
litt
in
g
cr
iter
ia
ar
e
ch
o
s
en
b
ased
o
n
s
o
m
e
q
u
ali
t
y
m
ea
s
u
r
es
s
u
c
h
as
in
f
o
r
m
a
tio
n
g
ai
n
,
w
h
ic
h
r
eq
u
ir
es
h
a
n
d
li
n
g
th
e
e
n
tire
d
ata
s
e
t
o
f
ea
ch
ex
p
a
n
d
in
g
n
o
d
es.
T
h
is
m
ak
e
s
it d
if
f
ic
u
lt
f
o
r
d
ec
is
io
n
tr
ee
s
to
b
e
ap
p
lied
to
b
ig
d
ata
ap
p
licatio
n
s
[
1
2
]
.
Su
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
(
SVM)
is
a
b
in
ar
y
clas
s
i
f
ier
w
h
ic
h
f
i
n
d
s
lin
ea
r
clas
s
i
f
ier
in
h
ig
h
e
r
d
i
m
en
s
io
n
al
f
ea
tu
r
e
s
p
ac
e
to
w
h
ic
h
o
r
ig
i
n
al
d
ata
s
p
ac
e
i
s
m
ap
p
ed
.
SVM
s
h
o
w
s
v
er
y
g
o
o
d
p
er
f
o
r
m
a
n
ce
f
o
r
d
ata
s
ets in
a
m
o
d
er
ate
s
ize.
I
t
h
as i
n
h
er
e
n
t li
m
itat
io
n
s
to
b
ig
d
ata
ap
p
licatio
n
s
[
1
2
]
.
Dee
p
m
ac
h
i
n
e
lear
n
i
n
g
h
as
b
ec
o
m
e
a
r
esear
c
h
f
r
o
n
tier
i
n
ar
tific
ial
i
n
telli
g
e
n
ce
.
I
t
is
a
m
ac
h
in
e
lear
n
in
g
tec
h
n
iq
u
e,
w
h
er
e
m
an
y
la
y
er
s
o
f
in
f
o
r
m
at
io
n
p
r
o
ce
s
s
in
g
s
ta
g
es
ar
e
ex
p
l
o
ited
in
h
ier
ar
ch
ica
l
ar
ch
itect
u
r
es.
I
t
co
m
p
u
tes
h
ier
ar
ch
ical
f
ea
tu
r
e
s
o
r
r
ep
r
esen
ta
tio
n
s
o
f
t
h
e
o
b
s
er
v
atio
n
al
d
ata
,
w
h
er
e
t
h
e
h
ig
h
er
-
lev
el
f
ea
t
u
r
es
o
r
f
ac
to
r
s
ar
e
d
ef
in
ed
f
r
o
m
lo
w
er
-
lev
e
l
o
n
e
s
.
Dee
p
lear
n
i
n
g
alg
o
r
it
h
m
s
ex
tr
ac
t
h
ig
h
-
le
v
el,
co
m
p
le
x
ab
s
tr
ac
tio
n
s
as
d
ata
r
ep
r
esen
tatio
n
s
th
r
o
u
g
h
a
h
ier
ar
ch
ical
lear
n
i
n
g
p
r
o
ce
s
s
.
W
h
il
e
d
ee
p
lea
r
n
in
g
ca
n
b
e
ap
p
lied
to
lear
n
f
r
o
m
la
b
eled
d
ata,
it
is
p
r
i
m
ar
il
y
attr
ac
tiv
e
f
o
r
lear
n
i
n
g
f
r
o
m
lar
g
e
a
m
o
u
n
ts
o
f
u
n
lab
eled
/
u
n
s
u
p
er
v
is
ed
d
ata,
m
ak
in
g
it
attr
ac
ti
v
e
f
o
r
ex
tr
ac
t
in
g
m
ea
n
in
g
f
u
l
r
ep
r
esen
tatio
n
s
an
d
p
atter
n
s
f
r
o
m
b
ig
d
ata.
Dee
p
lear
n
in
g
al
g
o
r
ith
m
s
a
n
d
ar
ch
itect
u
r
es
ar
e
m
o
r
e
ap
tl
y
s
u
ited
to
ad
d
r
ess
is
s
u
es
r
elate
d
to
Vo
lu
m
e
a
n
d
Var
iet
y
o
f
B
i
g
d
ata
an
al
y
tics
.
Dee
p
m
ac
h
i
n
e
lea
r
n
in
g
ca
n
b
e
ap
p
lied
to
b
ig
d
at
a.
Ho
w
ev
er
,
it
h
a
s
s
o
m
e
r
estrictio
n
s
i
n
b
ig
d
ata
a
p
p
licatio
n
s
b
ec
au
s
e
it r
eq
u
ir
e
s
s
ig
n
i
f
ican
t a
m
o
u
n
t o
f
tr
ain
i
n
g
ti
m
e
[
1
2
]
,
[
1
3
]
.
P
ar
allel
lear
n
er
f
o
r
ass
e
m
b
li
n
g
n
u
m
er
o
u
s
e
n
s
e
m
b
le
tr
ee
s
(
P
L
ANE
T
)
is
a
r
eg
r
ess
io
n
tr
ee
alg
o
r
ith
m
i
m
p
le
m
en
ted
w
it
h
a
s
eq
u
e
n
ce
o
f
Ma
p
R
ed
u
ce
j
o
b
s
th
at
r
u
n
o
n
t
h
e
b
ig
d
ata
f
r
a
m
e
w
o
r
k
,
H
ad
o
o
p
.
I
t
ca
n
d
ea
l
w
it
h
b
ig
v
o
lu
m
e
o
f
d
ata,
b
u
t is
n
o
t a
p
p
licab
le
to
d
ata
w
it
h
ca
t
eg
o
r
ical
attr
ib
u
tes [
1
2
]
.
On
e
tr
en
d
i
n
m
ac
h
in
e
lear
n
i
n
g
is
to
co
m
b
in
e
r
es
u
lt
s
o
f
m
u
ltip
le
lear
n
er
s
to
o
b
tain
b
etter
ac
cu
r
ac
y
.
T
h
is
tr
en
d
is
co
m
m
o
n
l
y
k
n
o
w
n
as
E
n
s
e
m
b
le
L
ea
r
n
i
n
g
.
T
h
er
e
ar
e
f
o
u
r
m
et
h
o
d
s
o
f
co
m
b
i
n
i
n
g
m
u
lt
ip
le
m
o
d
el
s
: b
ag
g
i
n
g
,
b
o
o
s
tin
g
,
s
ta
ck
in
g
,
an
d
er
r
o
r
-
co
r
r
ec
tin
g
o
u
t
p
u
t [
1
4
]
.
A
co
m
p
ar
i
s
o
n
o
f
s
e
v
er
al
m
a
ch
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
was
m
ad
e
i
n
T
ab
le
1
[
1
5
]
ac
co
r
d
in
g
to
alg
o
r
ith
m
s
t
y
p
e,
al
g
o
r
ith
m
s
t
r
ait,
lear
n
i
n
g
p
o
lic
y
,
lear
n
in
g
al
g
o
r
ith
m
s
,
a
n
d
cla
s
s
i
f
icati
o
n
s
tr
ate
g
y
.
So
m
e
f
ea
t
u
r
es o
f
m
ac
h
i
n
e
lear
n
in
g
a
lg
o
r
ith
m
s
w
er
e
co
m
p
ar
ed
in
T
ab
le
2
[
1
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
Ma
ch
in
e
Lea
r
n
in
g
in
B
ig
Da
t
a
(
Lid
o
n
g
W
a
n
g
)
119
T
ab
le
1
.
Su
m
m
ar
y
o
f
s
e
v
er
al
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
A
l
g
o
r
i
t
h
ms
A
l
g
o
r
i
t
h
ms
t
y
p
e
A
l
g
o
r
i
t
h
ms
c
h
a
r
a
c
t
e
r
i
st
i
c
L
e
a
r
n
i
n
g
p
o
l
i
c
y
L
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
C
l
a
ssi
f
i
c
a
t
i
o
n
st
r
a
t
e
g
y
D
e
c
i
si
o
n
t
r
e
e
D
i
scri
m
i
n
a
n
t
C
l
a
ssi
f
i
c
a
t
i
o
n
t
r
e
e
R
e
g
u
l
a
r
i
z
e
d
max
i
mu
m
l
i
k
e
l
i
h
o
o
d
e
st
i
mat
i
o
n
F
e
a
t
u
r
e
se
l
e
c
t
i
o
n
,
g
e
n
e
r
a
t
i
o
n
,
p
r
u
n
e
IF
-
T
H
EN
p
o
l
i
c
y
b
a
se
d
o
n
t
r
e
e
sp
i
t
t
i
n
g
N
o
n
-
l
i
n
e
a
r
su
p
p
o
r
t
v
e
c
t
o
r
mac
h
i
n
e
(
b
a
se
d
o
n
l
i
b
sv
m)
D
i
scri
m
i
n
a
n
t
S
u
p
e
r
-
p
l
a
n
e
se
p
a
r
a
t
i
o
n
,
k
e
r
n
e
l
t
r
i
c
k
M
i
n
i
m
i
z
i
n
g
t
h
e
l
o
ss o
f
r
e
g
u
l
a
r
h
i
n
g
e
,
so
f
t
marg
i
n
max
i
mi
z
a
t
i
o
n
S
e
q
u
e
n
t
i
a
l
mi
n
i
mal
o
p
t
i
m
i
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
(
S
M
O
)
M
a
x
i
m
u
m
c
l
a
ss o
f
t
e
st
sam
p
l
e
s
L
i
n
e
a
r
S
V
M
(
b
a
se
d
o
n
l
i
b
l
i
n
e
a
r
)
D
i
scri
m
i
n
a
n
t
S
u
p
e
r
-
p
l
a
n
e
se
p
a
r
a
t
i
o
n
M
i
n
i
m
i
z
i
n
g
t
h
e
l
o
ss o
f
r
e
g
u
l
a
r
h
i
n
g
e
,
so
f
t
marg
i
n
max
i
mi
z
a
t
i
o
n
S
e
q
u
e
n
t
i
a
l
d
u
a
l
me
t
h
o
d
M
a
x
i
m
u
m
w
e
i
g
h
t
e
d
t
e
st
samp
l
e
S
t
o
c
h
a
st
i
c
g
r
a
d
i
e
n
t
b
o
o
st
i
n
g
D
i
scri
m
i
n
a
n
t
L
i
n
e
a
r
c
o
mb
i
n
a
t
i
o
n
o
f
w
e
a
k
c
l
a
ss
i
f
i
e
r
(
b
a
se
d
o
n
d
e
c
i
si
o
n
t
r
e
e
)
A
d
d
i
t
i
o
n
mi
n
i
mi
z
a
t
i
o
n
l
o
ss
S
t
o
c
h
a
st
i
c
g
r
a
d
i
e
n
t
d
e
sce
n
t
a
l
g
o
r
i
t
h
m
L
i
n
e
a
r
c
o
mb
i
n
a
t
i
o
n
o
f
w
e
i
g
h
t
e
d
max
i
mu
m w
e
a
k
c
l
a
ssi
f
i
e
r
s
N
a
i
v
e
B
a
y
e
si
a
n
c
l
a
ssi
f
i
e
r
G
e
n
e
r
a
t
i
v
e
Jo
i
n
t
d
i
st
r
i
b
u
t
i
o
n
o
f
f
e
a
t
u
r
e
a
n
d
c
l
a
ss
,
c
o
n
d
i
t
i
o
n
a
l
i
n
d
e
p
e
n
d
e
n
t
a
ssu
m
p
t
i
o
n
M
a
x
i
m
u
m
l
i
k
e
l
i
h
o
o
d
e
st
i
mat
i
o
n
,
M
a
x
i
m
u
m
p
o
st
e
r
i
o
r
p
r
o
b
a
b
i
l
i
t
y
P
r
o
b
a
b
i
l
i
st
i
c
c
o
mp
u
t
a
t
i
o
n
M
a
x
i
m
u
m
p
o
st
e
r
i
o
r
p
r
o
b
a
b
i
l
i
t
y
T
ab
le
2
.
C
o
m
p
ar
in
g
m
ac
h
i
n
e
l
ea
r
n
in
g
al
g
o
r
ith
m
s
D
e
c
i
s
i
o
n
T
r
e
e
s
N
e
u
r
a
l
N
e
t
w
o
r
k
s
N
a
ï
v
e
B
a
y
e
s
k
N
N
S
V
M
R
u
l
e
-
l
e
a
r
n
e
r
s
A
c
c
u
r
a
c
y
i
n
g
e
n
e
r
a
l
**
***
*
**
****
**
S
p
e
e
d
o
f
l
e
a
r
n
i
n
g
w
i
t
h
r
e
sp
e
c
t
t
o
n
u
m
b
e
r
o
f
a
t
t
r
i
b
u
t
e
s
a
n
d
t
h
e
n
u
mb
e
r
o
f
i
n
st
a
n
c
e
s
***
*
****
****
*
**
S
p
e
e
d
o
f
c
l
a
ssi
f
i
c
a
t
i
o
n
****
****
****
*
****
****
T
o
l
e
r
a
n
c
e
t
o
mi
ss
i
n
g
v
a
l
u
e
s
***
*
****
*
**
**
T
o
l
e
r
a
n
c
e
t
o
i
r
r
e
l
e
v
a
n
t
a
t
t
r
i
b
u
t
e
s
***
*
**
**
****
**
T
o
l
e
r
a
n
c
e
t
o
r
e
d
u
n
d
a
n
t
a
t
t
r
i
b
u
t
e
s
**
**
*
**
****
**
T
o
l
e
r
a
n
c
e
t
o
h
i
g
h
l
y
i
n
t
e
r
d
e
p
e
n
d
e
n
t
a
t
t
r
i
b
u
t
e
s
(
e
.
g
.
p
a
r
i
t
y
p
r
o
b
l
e
ms)
**
***
*
*
***
**
D
e
a
l
i
n
g
w
i
t
h
d
i
s
c
r
e
t
e
/
b
i
n
a
r
y
/
c
o
n
t
i
n
u
o
u
s
a
t
t
r
i
b
u
t
e
s
****
*
*
*
(
n
o
t
d
i
s
c
r
e
t
e
)
***
(
n
o
t
c
o
n
t
i
n
u
o
u
s)
***
(
n
o
t
d
i
r
e
c
t
l
y
d
i
s
c
r
e
t
e
)
**
(
n
o
t
d
i
s
c
r
e
t
e
)
***
(
n
o
t
d
i
r
e
c
t
l
y
c
o
n
t
i
n
u
o
u
s)
T
o
l
e
r
a
n
c
e
t
o
n
o
i
se
**
**
***
*
**
*
D
e
a
l
i
n
g
w
i
t
h
d
a
n
g
e
r
o
f
o
v
e
r
f
i
t
t
i
n
g
**
*
***
***
**
**
A
t
t
e
mp
t
s fo
r
i
n
c
r
e
me
n
t
a
l
l
e
a
r
n
i
n
g
**
***
****
****
**
*
Ex
p
l
a
n
a
t
i
o
n
a
b
i
l
i
t
y
/
t
r
a
n
sp
a
r
e
n
c
y
o
f
k
n
o
w
l
e
d
g
e
/
c
l
a
ssi
f
i
c
a
t
i
o
n
s
****
*
****
**
*
****
M
o
d
e
l
p
a
r
a
me
t
e
r
h
a
n
d
l
i
n
g
***
*
****
***
*
***
(
*
*
*
*
s
tar
s
r
ep
r
esen
t t
h
e
b
est
an
d
*
s
tar
t
h
e
w
o
r
s
t p
er
f
o
r
m
a
n
ce
)
T
h
er
e
ar
e
s
ev
er
al
f
r
a
m
e
w
o
r
k
s
,
lik
e
Ma
p
/
R
ed
u
ce
,
Dr
y
ad
L
I
NQ
,
an
d
I
B
M
p
ar
allel
m
ac
h
i
n
e
lear
n
in
g
to
o
lb
o
x
th
at
h
a
v
e
ca
p
ab
ilit
ies t
o
s
ca
le
u
p
m
ac
h
in
e
lear
n
i
n
g
[
1
]
.
Ma
h
o
u
t
is
an
o
p
en
s
o
u
r
ce
m
a
ch
in
e
lear
n
in
g
l
ib
r
ar
y
f
r
o
m
Ap
ac
h
e
f
o
r
b
ig
d
ata
an
al
y
s
is
.
I
t
ai
m
s
to
b
e
th
e
m
ac
h
i
n
e
lear
n
i
n
g
to
o
l
o
f
ch
o
ice
w
h
e
n
th
e
co
llectio
n
o
f
d
ata
is
v
er
y
lar
g
e
[
4
]
.
T
h
e
A
p
ac
h
e
Ma
h
o
u
t
p
r
o
j
ec
t
ai
m
s
at
b
u
ild
in
g
a
s
ca
lab
le
m
a
ch
in
e
lear
n
in
g
l
ib
r
ar
y
o
n
to
p
o
f
Had
o
o
p
.
T
h
e
Ma
h
o
u
t
m
ac
h
in
e
l
ea
r
n
i
n
g
lib
r
ar
y
w
a
s
i
n
teg
r
ated
,
ad
ap
ted
,
an
d
ex
ten
d
ed
b
y
d
ev
elo
p
in
g
ad
v
a
n
ce
d
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
it
h
m
s
f
o
r
lar
g
e
s
ca
le
d
ata.
Ma
h
o
u
t
m
a
y
s
ig
n
i
f
ica
n
tl
y
h
elp
to
w
ar
d
s
g
r
o
u
p
in
g
s
i
m
ilar
ite
m
s
,
id
en
ti
f
y
in
g
m
ai
n
o
r
“
h
o
t”
to
p
ics,
ass
i
g
n
in
g
ite
m
s
to
p
r
ed
ef
in
ed
ca
t
eg
o
r
ies,
r
ec
o
m
m
e
n
d
in
g
i
m
p
o
r
ta
n
t
d
ata
to
d
iv
er
s
e
s
tak
e
h
o
ld
er
s
,
an
d
d
is
co
v
er
in
g
f
r
eq
u
e
n
t a
n
d
m
ea
n
in
g
f
u
l p
atter
n
s
i
n
a
s
p
ec
i
f
ic
d
ec
is
io
n
-
m
ak
i
n
g
s
ett
in
g
[
1
7
]
.
P
iv
o
talR
is
a
p
ac
k
a
g
e
f
o
r
m
a
ch
in
e
lear
n
i
n
g
o
n
b
ig
d
ata.
P
iv
o
talR
u
tili
ze
s
th
e
f
u
ll
p
o
w
er
o
f
p
ar
allel
co
m
p
u
tatio
n
a
n
d
d
is
tr
ib
u
t
iv
e
s
to
r
ag
e,
a
n
d
t
h
u
s
g
iv
e
s
t
h
e
n
o
r
m
al
R
u
s
er
ac
ce
s
s
to
b
ig
d
ata
s
to
r
ed
in
d
is
tr
ib
u
ti
v
e
d
atab
ases
o
r
Had
o
o
p
d
is
tr
ib
u
tiv
e
f
ile
s
y
s
te
m
(
HDFS)
.
I
t
p
r
o
v
id
es
d
ata
-
p
ar
al
lel
i
m
p
le
m
e
n
tatio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
1
7
–
1
2
3
120
o
f
m
ath
e
m
atica
l,
s
ta
tis
tica
l
a
n
d
m
ac
h
i
n
e
-
lear
n
i
n
g
al
g
o
r
ith
m
s
f
o
r
s
tr
u
ct
u
r
ed
an
d
u
n
s
tr
u
ctu
r
ed
d
ata.
T
h
u
s
P
iv
o
talR
also
en
ab
le
s
th
e
u
s
er
to
ap
p
ly
m
ac
h
i
n
e
lear
n
i
n
g
alg
o
r
ith
m
s
o
n
b
ig
d
ata
[
1
8
]
.
T
h
er
e
ar
e
a
lo
t
o
f
tech
n
o
lo
g
ies
s
u
p
p
o
r
tin
g
B
ig
Data
a
n
al
y
tic
s
an
d
ap
p
licatio
n
s
.
T
ab
le
3
[
1
1
]
co
m
p
ar
es
a
n
u
m
b
er
o
f
b
ig
d
ata
tech
n
o
lo
g
ie
s
.
T
h
e
tab
le
h
i
g
h
l
ig
h
t
s
t
h
e
d
if
f
er
en
t
t
y
p
e
s
o
f
s
y
s
te
m
s
a
n
d
th
ei
r
co
m
p
ar
ati
v
e
s
tr
e
n
g
th
s
an
d
w
e
ak
n
e
s
s
e
s
.
T
ab
le
3
.
C
o
m
p
ar
is
o
n
o
f
B
ig
D
ata
T
ec
h
n
o
lo
g
ies
In
-
me
mo
r
y
d
a
t
a
b
a
se
M
P
P
d
a
t
a
b
a
se
B
i
g
D
a
t
a
a
p
p
l
i
a
n
c
e
H
a
d
o
o
p
N
o
S
Q
L
d
a
t
a
b
a
se
C
o
n
si
s
t
e
n
t
W
W
W
P
P
A
v
a
i
l
a
b
l
e
W
W
W
P
P
F
a
u
l
t
t
o
l
e
r
a
n
t
W
W
P
W
W
S
u
i
t
a
b
l
e
f
o
r
r
e
a
l
-
t
i
me
t
r
a
n
sac
t
i
o
n
s
W
W
W
F
F
S
u
i
t
a
b
l
e
f
o
r
a
n
a
l
y
t
i
c
s
P
P
W
W
F
S
u
i
t
a
b
l
e
f
o
r
e
x
t
r
e
mel
y
b
i
g
d
a
t
a
F
P
P
W
W
S
u
i
t
a
b
l
e
f
o
r
u
n
st
r
u
c
t
u
r
e
d
d
a
t
a
F
F
P
W
W
W
: M
ee
ts
w
id
el
y
h
e
ld
ex
p
ec
ta
tio
n
s
.
P
: Po
ten
tiall
y
m
ee
t
s
w
id
el
y
h
e
ld
ex
p
ec
tatio
n
s
.
F: Fails
to
m
ee
t
w
id
el
y
h
eld
e
x
p
ec
tatio
n
s
.
3.
E
XAM
P
L
E
S O
F
M
ACH
I
NE
L
E
ARN
I
N
G
AP
P
L
I
CAT
I
O
NS IN
B
I
G
DATA
T
h
e
co
m
b
i
n
atio
n
o
f
s
u
p
er
v
i
s
ed
an
d
u
n
s
u
p
er
v
is
ed
m
ac
h
i
n
e
lear
n
in
g
tec
h
n
iq
u
e
s
f
o
r
ef
f
icien
tl
y
an
al
y
z
in
g
a
b
ig
v
o
l
u
m
e
o
f
cr
i
m
e
d
ata
w
a
s
p
r
o
p
o
s
ed
.
T
h
e
c
o
m
b
i
n
atio
n
in
cl
u
d
es
th
r
ee
s
te
p
s
:
d
i
m
en
s
io
n
ali
t
y
r
ed
u
ctio
n
,
clu
s
ter
i
n
g
,
an
d
class
if
ica
tio
n
.
R
s
tatis
t
ical
s
o
f
t
w
a
r
e
w
a
s
u
s
ed
b
ec
au
s
e
it
is
a
p
o
w
er
f
u
l
to
o
l
to
d
ea
l
w
it
h
b
ig
d
ata.
T
h
e
s
p
ec
if
ic
w
o
r
k
is
o
u
tli
n
ed
as
f
o
llo
w
s
[
1
9
]
:
1.
Me
asu
r
e
co
r
r
elatio
n
b
et
w
ee
n
cr
i
m
e
an
d
s
o
cial
attr
ib
u
tes.
T
h
is
m
eth
o
d
r
ed
u
ce
s
d
i
m
e
n
s
io
n
alit
y
o
f
th
e
cr
i
m
e
d
ata.
2.
Use
u
n
s
u
p
er
v
i
s
ed
m
ac
h
in
e
le
ar
n
in
g
tech
n
iq
u
e
to
d
iv
id
e
cr
i
m
e
d
ata
in
to
g
r
o
u
p
s
;
u
s
e
k
-
m
ea
n
s
clu
s
ter
i
n
g
alg
o
r
it
h
m
to
clu
s
ter
th
e
cr
i
m
e
d
ata
i
n
to
d
an
g
er
o
u
s
,
av
er
ag
e,
an
d
s
a
f
e
r
eg
io
n
s
.
3.
Use
s
u
p
er
v
is
ed
m
ac
h
i
n
e
lear
n
i
n
g
tec
h
n
iq
u
e
to
p
r
ed
ict
w
h
et
h
er
a
p
a
r
ticu
lar
r
e
g
io
n
is
d
an
g
e
r
o
u
s
o
r
s
af
e;
u
s
e
d
ec
is
io
n
tr
ee
class
if
i
ca
tio
n
alg
o
r
it
h
m
to
p
er
f
o
r
m
p
r
ed
ictio
n
s
.
An
al
y
s
i
s
an
d
m
i
n
in
g
o
f
s
o
ci
al
n
et
w
o
r
k
d
ata
f
o
r
s
o
ciet
y
is
s
u
es
w
a
s
co
n
d
u
cted
u
s
i
n
g
B
ig
Data
.
So
cial
d
ata
m
in
i
n
g
is
t
h
e
p
r
o
ce
s
s
o
f
an
a
l
y
zi
n
g
,
r
ep
r
esen
ti
n
g
as
w
e
ll
as
e
x
tr
ac
ti
n
g
ac
tio
n
ab
le
p
atter
n
s
f
r
o
m
s
o
cial
n
et
w
o
r
k
d
ata.
Ma
ch
in
e
lear
n
in
g
an
d
s
te
m
m
in
g
alg
o
r
it
h
m
s
w
er
e
u
s
ed
to
class
if
y
th
e
t
w
ee
t
s
.
T
w
ee
t
s
ar
e
o
f
ten
i
n
th
e
p
atter
n
o
f
b
ig
d
ata.
T
h
e
p
r
ed
ictin
g
f
ea
t
u
r
es
f
r
o
m
t
w
ee
t
s
w
er
e
ex
tr
ac
ted
f
r
o
m
a
co
llectio
n
o
f
t
w
ee
t
s
;
s
to
p
p
in
g
w
o
r
d
s
w
er
e
r
e
m
o
v
ed
;
an
d
all
k
e
y
w
o
r
d
s
w
er
e
s
elec
ted
.
A
s
t
w
ee
ts
ar
e
v
er
y
s
h
o
r
t
an
d
m
a
y
co
n
tain
i
n
co
m
p
lete
s
e
n
te
n
ce
s
,
th
e
m
ea
n
in
g
o
f
t
h
e
t
w
ee
ts
m
a
y
b
e
am
b
i
g
u
o
u
s
.
I
n
m
ac
h
in
e
lear
n
in
g
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
es
(
SVM)
ar
e
s
u
p
er
v
i
s
ed
m
o
d
el
s
w
i
th
r
el
at
ed
lear
n
in
g
alg
o
r
it
h
m
s
th
at
a
n
al
y
ze
all
th
e
d
ata
w
h
ic
h
ar
e
u
s
ed
f
o
r
class
i
f
ica
ti
o
n
o
f
th
e
t
w
ee
t
s
.
Ste
m
m
i
n
g
al
g
o
r
ith
m
u
s
es
a
p
r
e
-
p
r
o
ce
s
s
i
n
g
task
i
n
tex
t
m
i
n
i
n
g
an
d
ca
n
b
e
u
s
ed
as
a
co
m
m
o
n
r
eq
u
ir
e
m
en
t
o
f
n
at
u
r
al
la
n
g
u
ag
e
p
r
o
ce
s
s
in
g
f
u
n
c
tio
n
s
.
St
e
m
m
in
g
a
lg
o
r
it
h
m
w
a
s
u
s
ed
to
ex
tr
ac
t
th
e
m
ai
n
k
e
y
w
o
r
d
s
o
r
r
o
o
t
w
o
r
d
s
f
r
o
m
t
h
e
t
w
ee
t
s
.
T
h
e
s
te
m
m
in
g
alg
o
r
ith
m
ca
n
b
e
ap
p
lied
to
p
r
ed
ict
th
e
k
e
y
w
o
r
d
s
f
r
o
m
t
h
e
t
w
ee
t
s
.
A
ll th
e
k
e
y
w
o
r
d
s
ar
e
class
i
f
ied
b
y
t
h
e
SV
M
alg
o
r
ith
m
[
2
0
]
.
4.
CH
AL
L
E
N
G
E
S O
F
M
ACH
I
NE
L
E
A
RNIN
G
AP
P
L
I
CA
T
I
O
NS
I
N
B
I
G
DATA
Gen
er
al
c
h
alle
n
g
e
s
ab
o
u
t
m
a
ch
in
e
lear
n
i
n
g
ar
e:
(
1
)
d
esi
g
n
in
g
s
ca
lab
le
a
n
d
f
le
x
ib
le
co
m
p
u
tatio
n
a
l
ar
ch
itect
u
r
es
f
o
r
m
ac
h
i
n
e
lear
n
in
g
;
(
2
)
th
e
ab
ilit
y
to
u
n
d
er
s
tan
d
ch
ar
ac
ter
is
tic
s
o
f
d
ata
b
ef
o
r
e
ap
p
ly
in
g
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
a
n
d
to
o
ls
;
an
d
(
3
)
th
e
ab
ilit
y
to
co
n
s
tr
u
ct,
lear
n
an
d
i
n
f
er
w
i
th
in
cr
ea
s
i
n
g
s
a
m
p
l
e
s
ize,
d
i
m
e
n
s
io
n
a
lit
y
,
a
n
d
ca
te
g
o
r
ies o
f
lab
els [
2
1
]
.
T
h
er
e
ar
e
m
an
y
s
ca
le
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
ith
m
s
,
b
u
t
m
a
n
y
i
m
p
o
r
tan
t
s
p
ec
if
ic
s
u
b
-
f
iel
d
s
in
lar
g
e
-
s
ca
le
m
ac
h
in
e
lear
n
in
g
,
s
u
c
h
as
lar
g
e
-
s
ca
le
r
ec
o
m
m
e
n
d
er
s
y
s
te
m
s
,
n
atu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
,
as
s
o
ciatio
n
r
u
le
lear
n
i
n
g
,
e
n
s
e
m
b
le
lear
n
i
n
g
,
s
till
f
ac
e
th
e
s
ca
lab
ilit
y
p
r
o
b
lem
s
[
1
].
T
h
e
b
asic
Ma
p
R
ed
u
ce
f
r
a
m
e
w
o
r
k
co
m
m
o
n
l
y
p
r
o
v
id
ed
b
y
f
ir
s
t
-
g
e
n
er
atio
n
“
B
i
g
Data
an
al
y
tics
”
p
latf
o
r
m
s
lik
e
Had
o
o
p
lack
s
an
ess
e
n
tia
l
f
ea
t
u
r
e
f
o
r
m
ac
h
i
n
e
lear
n
i
n
g
(
ML
)
.
Ma
p
R
ed
u
c
e
d
o
es
n
o
t
s
u
p
p
o
r
t
iter
atio
n
/r
ec
u
r
s
io
n
o
r
ce
r
tain
k
e
y
f
ea
t
u
r
es
r
eq
u
ir
ed
to
ef
f
i
cien
tl
y
iter
ate
“
ar
o
u
n
d
”
a
Ma
p
R
ed
u
ce
p
r
o
g
r
a
m
.
P
r
o
g
r
am
m
er
s
b
u
ild
in
g
M
L
m
o
d
els
o
n
s
u
c
h
s
y
s
te
m
s
ar
e
f
o
r
ce
d
to
im
p
le
m
e
n
t
lo
o
p
in
g
in
a
d
-
h
o
c
w
a
y
s
o
u
ts
id
e
th
e
co
r
e
Ma
p
R
ed
u
ce
f
r
a
m
e
wo
r
k
.
T
h
is
m
a
k
es
t
h
eir
p
r
o
g
r
am
m
in
g
ta
s
k
m
u
c
h
h
ar
d
er
,
an
d
it
o
f
ten
a
ls
o
y
ield
s
in
e
f
f
ic
ien
t
p
r
o
g
r
a
m
s
in
t
h
e
e
n
d
.
T
h
is
lack
o
f
s
u
p
p
o
r
t
h
as
m
o
tiv
ated
t
h
e
r
ec
en
t
d
ev
elo
p
m
e
n
t
o
f
v
ar
io
u
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
Ma
ch
in
e
Lea
r
n
in
g
in
B
ig
Da
t
a
(
Lid
o
n
g
W
a
n
g
)
121
s
p
ec
ia
lized
ap
p
r
o
ac
h
es
o
r
lib
r
ar
ies
to
s
u
p
p
o
r
t
iter
ativ
e
p
r
o
g
r
a
m
m
i
n
g
o
n
lar
g
e
clu
s
ter
s
.
Me
an
w
h
ile,
r
ec
en
t
Ma
p
R
ed
u
ce
e
x
ten
s
io
n
s
s
u
c
h
a
s
Ha
L
o
o
p
,
T
w
i
s
ter
,
an
d
P
r
I
tr
ai
m
at
d
ir
ec
tl
y
ad
d
r
ess
in
g
th
e
iter
atio
n
o
u
ta
g
e
i
n
Ma
p
R
ed
u
ce
; th
e
y
d
o
s
o
at
th
e
p
h
y
s
ical
lev
e
l [
2
2
]
.
Ma
j
o
r
p
r
o
b
lem
s
t
h
at
m
a
k
e
th
e
m
ac
h
i
n
e
lear
n
i
n
g
(
M
L
)
tech
n
iq
u
es
u
n
s
u
i
tab
le
f
o
r
s
o
lv
in
g
b
ig
d
ata
class
i
f
icatio
n
p
r
o
b
lem
s
ar
e:
(
1
)
A
n
M
L
tec
h
n
iq
u
e
t
h
at
is
tr
ain
ed
o
n
a
p
ar
ticu
lar
lab
ele
d
d
atasets
o
r
d
ata
d
o
m
ai
n
m
a
y
n
o
t
b
e
s
u
i
tab
le
f
o
r
an
o
th
er
d
ataset
o
r
d
ata
d
o
m
ai
n
–
th
a
t
th
e
clas
s
i
f
icatio
n
m
a
y
n
o
t
b
e
r
o
b
u
s
t
o
v
er
d
if
f
er
en
t
d
ataset
s
o
r
d
ata
d
o
m
ain
s
;
(
2
)
an
ML
tech
n
iq
u
e
is
in
g
en
er
al
tr
ain
ed
u
s
i
n
g
a
ce
r
tain
n
u
m
b
er
o
f
class
t
y
p
es
an
d
h
e
n
ce
a
lar
g
e
v
ar
ietie
s
o
f
cla
s
s
t
y
p
es
f
o
u
n
d
in
a
d
y
n
a
m
icall
y
g
r
o
w
in
g
d
ataset
w
ill
lead
to
in
ac
cu
r
ate
clas
s
i
f
icatio
n
r
es
u
lt
s
;
a
n
d
(
3
)
an
M
L
tec
h
n
iq
u
e
i
s
d
ev
elo
p
ed
b
ased
o
n
a
s
i
n
g
le
lear
n
in
g
tas
k
,
a
n
d
th
u
s
th
e
y
ar
e
n
o
t
s
u
itab
le
f
o
r
t
o
d
ay
’
s
m
u
ltip
le
lear
n
i
n
g
ta
s
k
s
an
d
k
n
o
w
led
g
e
tr
a
n
s
f
er
r
eq
u
i
r
e
m
en
t
s
o
f
B
ig
d
ata
an
al
y
tics
[
2
3
]
.
T
r
a
d
itio
n
al
alg
o
r
ith
m
s
i
n
m
ac
h
in
e
lear
n
i
n
g
g
en
er
all
y
d
o
n
o
t
s
ca
le
to
b
ig
d
ata.
T
h
e
m
ai
n
d
if
f
ic
u
lt
y
lies
w
i
th
th
eir
m
e
m
o
r
y
co
n
s
tr
ain
t.
A
l
th
o
u
g
h
al
g
o
r
ith
m
s
t
y
p
icall
y
a
s
s
u
m
e
t
h
at
tr
ai
n
i
n
g
d
ata
s
a
m
p
les
e
x
is
t
i
n
m
ai
n
m
e
m
o
r
y
,
b
i
g
d
ata
d
o
es
n
o
t
f
it
i
n
to
it.
A
co
m
m
o
n
a
p
p
r
o
ac
h
to
lear
n
in
g
f
r
o
m
a
l
ar
g
e
d
ataset
is
d
ata
d
is
tr
ib
u
tio
n
.
B
y
r
ep
laci
n
g
b
at
ch
tr
ain
in
g
o
n
t
h
e
o
r
ig
i
n
al
tr
a
in
i
n
g
d
ata
s
et
w
it
h
s
ep
ar
ated
co
m
p
u
tatio
n
s
o
n
th
e
d
is
tr
ib
u
ted
s
u
b
s
et
s
,
o
n
e
ca
n
tr
ain
a
n
al
ter
n
ati
v
e
p
r
ed
ictio
n
m
o
d
el
at
a
s
ac
r
i
f
ice
o
f
ac
c
u
r
ac
y
.
An
o
th
er
w
a
y
is
to
u
s
e
o
n
li
n
e
lear
n
in
g
,
i
n
w
h
ic
h
m
e
m
o
r
y
u
s
a
g
e
d
o
es
n
o
t
d
ep
en
d
o
n
t
h
e
s
ize
o
f
th
e
d
at
aset.
Neit
h
er
o
n
li
n
e
lear
n
in
g
n
o
r
d
is
tr
ib
u
ted
lear
n
i
n
g
i
s
s
u
f
f
icie
n
t
f
o
r
lear
n
in
g
f
r
o
m
b
i
g
d
ata
s
tr
ea
m
s
.
T
h
er
e
ar
e
t
w
o
r
ea
s
o
n
s
.
F
ir
s
t
is
t
h
at
t
h
e
d
ata
s
ize
is
to
o
la
r
g
e
to
b
e
r
elax
ed
b
y
eit
h
er
o
n
li
n
e
o
r
d
is
tr
ib
u
ted
lear
n
in
g
.
Seq
u
en
tial
o
n
lin
e
lear
n
in
g
o
n
b
i
g
d
ata
r
eq
u
ir
es
t
o
o
m
u
c
h
ti
m
e
f
o
r
tr
ai
n
in
g
o
n
a
s
i
n
g
le
m
ac
h
in
e.
O
n
t
h
e
o
t
h
e
r
h
a
n
d
,
d
is
tr
ib
u
ted
lear
n
in
g
w
it
h
a
lar
g
e
n
u
m
b
er
o
f
m
ac
h
i
n
e
s
r
ed
u
ce
s
t
h
e
g
ai
n
ed
ef
f
ici
e
n
c
y
p
er
m
ac
h
i
n
e
a
n
d
af
f
ec
ts
th
e
o
v
er
all
p
er
f
o
r
m
a
n
ce
.
T
h
e
s
ec
o
n
d
r
ea
s
o
n
is
th
a
t
co
m
b
i
n
i
n
g
r
ea
l
-
ti
m
e
tr
ai
n
in
g
a
n
d
p
r
ed
ictio
n
h
as
n
o
t
b
ee
n
s
t
u
d
ied
.
Sin
ce
b
ig
d
ata
is
t
y
p
icall
y
u
s
e
d
af
ter
b
ein
g
s
to
r
ed
in
(
d
is
tr
ib
u
ted
)
s
to
r
ag
e,
th
e
lear
n
i
n
g
p
r
o
ce
s
s
also
ten
d
s
to
w
o
r
k
i
n
a
b
atch
m
an
n
er
[
2
4
]
.
Scalin
g
u
p
b
ig
d
ata
to
p
r
o
p
er
d
i
m
e
n
s
io
n
alit
y
is
a
c
h
alle
n
g
e
th
at
ca
n
e
n
co
u
n
ter
i
n
m
ac
h
i
n
e
lear
n
i
n
g
alg
o
r
ith
m
s
; a
n
d
t
h
er
e
ar
e
ch
all
en
g
e
s
o
f
d
ea
li
n
g
w
i
th
v
elo
cit
y
,
v
o
lu
m
e
an
d
m
an
y
m
o
r
e
f
o
r
all
t
y
p
es o
f
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
i
t
h
m
s
.
Si
n
ce
b
ig
d
ata
p
r
o
ce
s
s
in
g
r
eq
u
ir
es
d
ec
o
m
p
o
s
itio
n
,
p
ar
allelis
m
,
m
o
d
u
lar
it
y
a
n
d
/o
r
r
ec
u
r
r
en
ce
,
in
f
le
x
ib
le
b
lack
-
b
o
x
t
y
p
e
m
ac
h
i
n
e
lear
n
i
n
g
m
o
d
els f
ailed
i
n
an
o
u
ts
et
[
2
]
.
A
p
p
l
y
in
g
th
e
d
is
tr
ib
u
ted
d
ata
-
p
ar
allelis
m
(
DDP
)
p
atter
n
s
in
B
ig
Data
B
a
y
esia
n
Net
w
o
r
k
(
B
N)
lear
n
in
g
f
ac
e
s
s
e
v
er
al
c
h
alle
n
g
es:
(
1
)
ef
f
ec
ti
v
el
y
p
r
e
-
p
r
o
ce
s
s
in
g
b
i
g
d
ata
to
ev
al
u
ate
its
q
u
alit
y
an
d
r
ed
u
ce
th
e
s
ize
i
f
n
ec
es
s
ar
y
;
(
2
)
d
esig
n
i
n
g
a
w
o
r
k
f
lo
w
ca
p
ab
le
o
f
tak
in
g
Gi
g
ab
y
te
s
o
f
b
i
g
d
ata
s
et
s
a
n
d
lear
n
i
n
g
B
N
s
w
it
h
d
ec
en
t a
cc
u
r
ac
y
; (
3
)
p
r
o
v
id
in
g
ea
s
y
s
ca
lab
ilit
y
s
u
p
p
o
r
t t
o
B
N
lear
n
in
g
al
g
o
r
ith
m
s
[
1
4
]
.
Dee
p
lear
n
in
g
c
h
alle
n
g
e
s
in
b
ig
d
ata
an
al
y
tic
s
lie
in
:
in
cr
e
m
en
tal
lear
n
i
n
g
f
o
r
n
o
n
-
s
tati
o
n
ar
y
d
ata,
h
ig
h
-
d
i
m
en
s
io
n
al
d
ata,
an
d
lar
g
e
-
s
ca
le
m
o
d
el
s
[
1
3
]
.
B
ec
au
s
e
h
ig
h
-
le
v
el
d
ata
p
ar
all
el
f
r
a
m
e
w
o
r
k
s
,
li
k
e
Ma
p
R
ed
u
ce
d
o
n
o
t
n
a
tu
r
all
y
o
r
e
f
f
icie
n
t
l
y
s
u
p
p
o
r
t
m
an
y
i
m
p
o
r
tan
t d
ata
m
i
n
i
n
g
an
d
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
a
n
d
ca
n
lead
to
in
e
f
f
icien
t
lear
n
in
g
s
y
s
te
m
s
,
th
e
Gr
ap
h
L
ab
ab
s
tr
ac
tio
n
w
as
i
n
tr
o
d
u
ce
d
.
I
t
n
atu
r
all
y
e
x
p
r
ess
e
s
as
y
n
c
h
r
o
n
o
u
s
,
d
y
n
a
m
ic,
g
r
ap
h
-
p
ar
allel
co
m
p
u
tatio
n
w
h
ile
en
s
u
r
in
g
d
ata
co
n
s
is
t
en
c
y
a
n
d
ac
h
iev
in
g
a
h
i
g
h
d
eg
r
ee
o
f
p
ar
allel
p
er
f
o
r
m
a
n
ce
i
n
th
e
s
h
ar
ed
-
m
e
m
o
r
y
s
ett
in
g
[
2
5
]
.
5.
T
E
CH
NO
L
O
G
Y
P
RO
G
R
E
SS
O
F
M
ACH
I
NE
L
E
A
RNI
NG
AP
P
L
I
CA
T
I
O
N
S IN
B
I
G
DA
T
A
Mo
s
t
o
f
th
e
ad
v
an
ce
s
f
o
r
s
ca
lab
le
m
ac
h
i
n
e
lear
n
i
n
g
(
e.
g
.
Ma
d
lib
,
A
p
ac
h
e
Ma
h
o
u
t,
etc.
)
ar
e
h
ap
p
en
in
g
i
n
th
e
m
a
s
s
i
v
el
y
p
ar
allel
d
atab
ase
p
r
o
ce
s
s
in
g
co
m
m
u
n
it
y
.
B
etter
w
o
r
k
ca
n
b
e
d
o
n
e
in
t
h
e
B
ig
Data
er
a
b
y
d
esi
g
n
in
g
an
d
i
m
p
le
m
e
n
ti
n
g
m
ac
h
i
n
e
lear
n
i
n
g
al
g
o
r
it
h
m
s
w
it
h
s
ca
le
-
f
r
ie
n
d
l
y
p
r
ed
ictiv
e
f
u
n
ctio
n
s
.
T
h
e
f
o
llo
w
in
g
m
et
h
o
d
s
h
a
v
e
b
ee
n
ex
p
lo
r
in
g
a
n
d
ev
al
u
ati
n
g
[
2
1
]
:
(
1
)
d
ee
p
lear
n
in
g
al
g
o
r
ith
m
s
th
at
a
u
to
m
a
te
th
e
f
ea
t
u
r
e
en
g
i
n
ee
r
i
n
g
p
r
o
ce
s
s
b
y
lear
n
in
g
to
cr
ea
te
an
d
s
if
t
th
r
o
u
g
h
d
ata
-
d
r
iv
e
n
f
ea
t
u
r
es,
(
2
)
in
cr
e
m
en
tal
lear
n
in
g
al
g
o
r
ith
m
s
i
n
as
s
o
cia
tiv
e
m
e
m
o
r
y
ar
c
h
itect
u
r
es
t
h
at
ca
n
s
ea
m
le
s
s
l
y
ad
ap
t
to
f
u
tu
r
e
d
ata
s
a
m
p
le
s
a
n
d
s
o
u
r
ce
s
,
(
3
)
f
ac
eted
lear
n
in
g
t
h
at
ca
n
lear
n
h
ier
ar
c
h
ical
s
tr
u
ctu
r
e
in
th
e
d
ata,
an
d
(
4
)
m
u
lt
i
-
tas
k
lear
n
i
n
g
th
a
t
ca
n
lear
n
s
e
v
er
al
p
r
ed
ictiv
e
f
u
n
ctio
n
s
in
p
ar
allel.
T
h
e
B
ig
Data
class
i
f
icatio
n
r
eq
u
ir
es
m
u
lti
-
d
o
m
ai
n
,
r
ep
r
esen
tat
io
n
-
lear
n
i
n
g
(
MD
R
L
)
tech
n
iq
u
e
b
ec
au
s
e
o
f
it
s
lar
g
e
a
n
d
g
r
o
w
i
n
g
d
ata
d
o
m
ain
.
T
h
e
MD
R
L
tec
h
n
iq
u
e
i
n
cl
u
d
es
f
ea
t
u
r
e
v
ar
iab
le
lear
n
i
n
g
,
f
ea
t
u
r
e
ex
tr
ac
tio
n
lear
n
in
g
,
an
d
d
is
tan
ce
-
m
etr
ic
lear
n
i
n
g
.
Sev
er
al
r
ep
r
esen
tatio
n
-
lear
n
i
n
g
tech
n
iq
u
e
s
h
a
v
e
b
ee
n
p
r
o
p
o
s
ed
in
m
ac
h
i
n
e
lea
r
n
in
g
.
T
h
e
cr
o
s
s
-
d
o
m
ai
n
,
r
ep
r
esen
tat
io
n
-
lear
n
i
n
g
(
C
D
R
L
)
tech
n
iq
u
e
is
m
a
y
b
e
s
u
itab
le
f
o
r
th
e
B
ig
Data
cla
s
s
if
icatio
n
alo
n
g
w
it
h
th
e
s
u
g
g
es
ted
n
et
w
o
r
k
m
o
d
el
[
2
3
]
.
A
k
e
y
b
en
e
f
it
o
f
d
ee
p
lear
n
in
g
is
th
e
an
al
y
s
i
s
an
d
lear
n
i
n
g
o
f
m
a
s
s
i
v
e
a
m
o
u
n
ts
o
f
u
n
s
u
p
er
v
is
ed
d
ata,
m
ak
in
g
it
a
v
al
u
ab
le
to
o
l
f
o
r
B
ig
Data
a
n
al
y
tics
.
Ho
w
d
ee
p
lear
n
in
g
ca
n
b
e
u
tili
ze
d
i
n
B
ig
Data
a
n
al
y
tics
w
as
ex
p
lo
r
ed
;
th
is
i
n
cl
u
d
es
ex
tr
ac
tin
g
co
m
p
le
x
p
atter
n
s
f
r
o
m
m
ass
i
v
e
v
o
l
u
m
es
o
f
d
ata,
s
e
m
a
n
tic
i
n
d
ex
i
n
g
,
d
ata
tag
g
in
g
,
f
a
s
t
in
f
o
r
m
a
tio
n
r
etr
iev
al,
an
d
s
i
m
p
li
f
y
in
g
d
is
cr
i
m
i
n
ati
v
e
tas
k
s
.
So
m
e
f
u
r
th
er
r
esear
ch
o
f
d
ee
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
1
7
–
1
2
3
122
lear
n
in
g
in
B
ig
Data
w
as
also
in
v
e
s
ti
g
ated
;
th
i
s
in
cl
u
d
es
s
tr
e
a
m
i
n
g
d
ata,
h
ig
h
-
d
i
m
en
s
io
n
a
l
d
ata,
s
ca
lab
ilit
y
o
f
Dee
p
L
ea
r
n
i
n
g
m
o
d
els,
an
d
d
i
s
tr
ib
u
ted
co
m
p
u
t
in
g
[
1
3
]
.
As
an
i
m
p
o
r
ta
n
t
m
ac
h
i
n
e
lear
n
in
g
tech
n
iq
u
e,
B
ay
e
s
ia
n
Net
w
o
r
k
(
B
N)
h
as
b
ee
n
w
id
el
y
u
s
ed
to
m
o
d
el
p
r
o
b
ab
ilis
tic
r
elatio
n
s
h
ip
s
a
m
o
n
g
v
ar
iab
les.
An
in
te
llig
e
n
t
B
ig
Data
p
r
e
-
p
r
o
ce
s
s
i
n
g
ap
p
r
o
ac
h
an
d
a
d
ata
q
u
alit
y
s
co
r
e
w
er
e
p
r
o
p
o
s
ed
to
m
ea
s
u
r
e
a
n
d
en
s
u
r
e
th
e
d
ata
q
u
alit
y
an
d
d
ata
f
ai
th
f
u
l
n
es
s
;
a
n
e
w
w
ei
g
h
t
b
ased
en
s
e
m
b
le
al
g
o
r
ith
m
w
a
s
p
r
o
p
o
s
ed
t
o
lear
n
a
B
N
s
tr
u
ctu
r
e
f
r
o
m
a
n
en
s
e
m
b
le
o
f
lo
ca
l
r
esu
lt
s
.
Fo
r
ea
s
il
y
in
te
g
r
atin
g
t
h
e
alg
o
r
it
h
m
w
i
th
d
is
tr
ib
u
ted
d
ata
-
p
ar
allelis
m
(
DDP
)
en
g
in
e
s
,
s
u
c
h
as
Had
o
o
p
,
Kep
ler
s
cien
tif
i
c
w
o
r
k
f
lo
w
w
as
e
m
p
lo
y
ed
to
b
u
ild
th
e
w
h
o
le
lear
n
i
n
g
p
r
o
ce
s
s
.
Ho
w
Kep
ler
ca
n
f
ac
ili
tate
b
u
ild
in
g
a
n
d
r
u
n
n
i
n
g
th
e
B
ig
Data
B
N
lear
n
i
n
g
a
p
p
licatio
n
w
as
a
ls
o
d
e
m
o
n
s
t
r
ated
.
A
Sca
lab
le
B
a
y
esia
n
Net
w
o
r
k
L
ea
r
n
i
n
g
(
SB
NL
)
w
o
r
k
f
lo
w
w
a
s
d
esi
g
n
ed
th
r
o
u
g
h
co
m
b
i
n
in
g
m
ac
h
in
e
lear
n
in
g
,
d
is
tr
ib
u
ted
co
m
p
u
t
in
g
,
a
n
d
w
o
r
k
f
lo
w
tech
n
iq
u
es.
T
h
e
w
o
r
k
f
lo
w
i
n
c
lu
d
es
i
n
telli
g
e
n
t
B
ig
Da
ta
p
r
e
-
p
r
o
ce
s
s
i
n
g
a
n
d
ef
f
ec
ti
v
e
B
N
lear
n
in
g
f
r
o
m
B
ig
Data
b
y
le
v
er
a
g
in
g
e
n
s
e
m
b
le
l
ea
r
n
in
g
an
d
d
is
tr
ib
u
ted
co
m
p
u
tin
g
m
o
d
el
[
1
4
]
.
Fo
r
s
tr
ea
m
p
r
o
ce
s
s
i
n
g
,
o
n
e
m
u
s
t
p
r
o
ce
s
s
n
e
w
d
ata
in
r
ea
l
-
ti
m
e
an
d
in
m
a
n
y
t
i
m
e
s
,
co
n
s
id
er
s
th
e
h
is
to
r
ical
d
ata
as
w
ell
to
g
e
n
er
ate
a
v
alu
e.
Mo
s
t
o
f
te
n
,
s
tr
ea
m
p
r
o
ce
s
s
in
g
i
n
v
o
l
v
es
t
h
e
u
s
e
o
f
p
r
ev
io
u
s
l
y
tr
ain
ed
m
o
d
els
to
av
o
id
to
o
m
u
ch
p
r
o
ce
s
s
i
n
g
an
d
u
l
ti
m
atel
y
r
ed
u
ce
r
esp
o
n
s
e
tim
e
s
.
A
n
o
v
el
ar
ch
itect
u
r
e
f
o
r
p
er
f
o
r
m
in
g
m
ac
h
i
n
e
lear
n
in
g
o
v
er
b
i
g
d
ata
s
tr
ea
m
s
was
p
r
o
p
o
s
ed
.
T
h
e
ar
ch
itect
u
r
e
p
r
o
v
id
es
r
eliab
le
p
er
s
is
ten
t
s
to
r
ag
e
o
f
d
ata
o
v
er
th
e
Had
o
o
p
Dis
t
r
ib
u
ted
Fil
e
S
y
s
te
m
(
HDF
S)
an
d
HB
ase.
T
h
e
co
r
e
o
f
th
e
ar
ch
itect
u
r
e
is
co
m
p
r
is
ed
o
f
t
h
e
b
atch
-
an
d
s
tr
ea
m
-
p
r
o
ce
s
s
i
n
g
m
o
d
u
le
s
.
I
t
p
r
o
v
id
es
m
ac
h
in
e
lear
n
i
n
g
to
o
ls
an
d
alg
o
r
it
h
m
s
s
o
th
at
d
e
v
elo
p
er
s
ca
n
ea
s
il
y
ta
k
e
ad
v
an
ta
g
e
o
f
t
h
e
m
to
ca
r
r
y
o
u
t
tas
k
s
s
u
c
h
as
p
r
ed
ictio
n
,
clu
s
ter
i
n
g
,
r
ec
o
m
m
e
n
d
atio
n
,
a
n
d
class
i
f
icat
io
n
,
etc.
,
[
2
6
]
.
A
d
is
tr
ib
u
ted
s
tr
ea
m
in
g
a
lg
o
r
ith
m
w
a
s
p
r
o
p
o
s
ed
to
lear
n
d
ec
is
io
n
r
u
les
f
o
r
r
e
g
r
ess
io
n
task
s
.
T
h
e
alg
o
r
ith
m
i
s
a
v
ailab
le
i
n
Scal
ab
le
A
d
v
an
ce
d
Ma
s
s
i
v
e
O
n
li
n
e
An
al
y
s
i
s
(
S
A
MO
A
)
,
a
n
o
p
en
-
s
o
u
r
ce
p
latf
o
r
m
f
o
r
m
i
n
i
n
g
b
i
g
d
ata
s
tr
ea
m
s
.
I
t
u
s
es
a
h
y
b
r
id
o
f
v
er
t
ical
a
n
d
h
o
r
izo
n
tal
p
ar
allelis
m
to
d
is
tr
ib
u
te
A
d
ap
tiv
e
Mo
d
el
R
u
le
s
(
A
MR
u
les)
o
n
a
clu
s
ter
.
T
h
e
d
ec
is
io
n
r
u
le
s
b
u
ilt
b
y
AM
R
u
les
ar
e
co
m
p
r
eh
en
s
ib
le
m
o
d
els.
S
A
MO
A
is
a
f
r
a
m
e
w
o
r
k
t
h
at
ea
s
es
th
e
d
ev
e
lo
p
m
en
t
o
f
n
e
w
d
is
tr
ib
u
ted
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
a
n
d
th
e
d
ep
lo
y
m
en
t
o
f
th
e
s
e
i
m
p
le
m
en
tatio
n
s
o
n
to
p
o
f
s
tate
-
o
f
t
h
e
-
ar
t
d
is
tr
ib
u
ted
s
tr
ea
m
p
r
o
ce
s
s
in
g
e
n
g
in
e
s
(
DSP
E
s
)
.
I
t
is
also
a
lib
r
ar
y
o
f
d
is
tr
ib
u
ted
d
ata
m
i
n
i
n
g
a
n
d
m
ac
h
in
e
lear
n
in
g
al
g
o
r
ith
m
s
t
h
at
allo
w
s
u
s
er
s
to
u
s
e
o
r
cu
s
to
m
ize
e
x
is
ti
n
g
o
n
es
[
2
7
]
.
Featu
r
e
s
e
lectio
n
(
F
S)
is
a
n
i
m
p
o
r
ta
n
t
to
p
ic
i
n
m
ac
h
i
n
e
lea
r
n
in
g
a
n
d
d
ata
m
i
n
i
n
g
.
T
h
e
o
b
j
ec
tiv
e
o
f
f
ea
t
u
r
e
s
elec
tio
n
is
to
s
elec
t
a
s
u
b
s
et
o
f
r
ele
v
an
t
f
ea
tu
r
es
f
o
r
b
u
ild
in
g
e
f
f
ec
tiv
e
p
r
ed
ictio
n
m
o
d
el
s
.
Var
io
u
s
F
S
m
et
h
o
d
s
h
a
v
e
b
ee
n
p
r
o
p
o
s
ed
.
B
ased
o
n
th
e
s
e
lectio
n
cr
iter
i
o
n
ch
o
ice,
t
h
ese
m
et
h
o
d
s
ca
n
b
e
r
o
u
g
h
l
y
d
i
v
id
ed
in
to
th
r
ee
ca
te
g
o
r
ies:
f
ilter
m
eth
o
d
s
,
w
r
ap
p
er
m
et
h
o
d
s
,
an
d
em
b
ed
d
ed
m
et
h
o
d
s
ap
p
r
o
ac
h
es.
Fi
lter
m
et
h
o
d
s
r
elies
o
n
th
e
c
h
ar
ac
ter
is
t
ics
o
f
th
e
d
ata
s
u
c
h
as
co
r
r
elatio
n
,
d
is
tan
ce
a
n
d
in
f
o
r
m
atio
n
,
w
it
h
o
u
t
i
n
v
o
lv
i
n
g
an
y
lear
n
in
g
alg
o
r
it
h
m
.
W
r
ap
p
er
m
et
h
o
d
s
r
eq
u
ir
e
o
n
e
p
r
ed
eter
m
i
n
ed
lear
n
i
n
g
al
g
o
r
ith
m
f
o
r
e
v
al
u
ati
n
g
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
s
elec
ted
f
ea
tu
r
es
s
et.
E
m
b
ed
d
ed
m
et
h
o
d
s
ai
m
to
in
te
g
r
ate
t
h
e
f
ea
t
u
r
e
s
e
l
ec
tio
n
p
r
o
ce
s
s
i
n
to
th
e
m
o
d
el
tr
ai
n
in
g
p
r
o
ce
s
s
;
t
h
e
y
ar
e
f
as
ter
th
a
n
th
e
w
r
ap
p
er
m
et
h
o
d
s
;
an
d
s
ti
ll
p
r
o
v
id
e
s
u
i
t
ab
le
f
ea
tu
r
e
s
u
b
s
et
f
o
r
th
e
lear
n
i
n
g
a
lg
o
r
it
h
m
.
On
lin
e
f
ea
t
u
r
e
s
elec
tio
n
(
OF
S)
f
o
r
m
i
n
in
g
b
i
g
d
ata
w
as
s
t
u
d
ied
to
s
o
lv
e
t
h
e
f
ea
t
u
r
e
s
elec
tio
n
p
r
o
b
lem
b
y
a
n
o
n
li
n
e
lear
n
i
n
g
ap
p
r
o
ac
h
.
T
h
e
g
o
al
o
f
o
n
lin
e
f
ea
tu
r
e
s
elec
tio
n
w
a
s
to
d
ev
elo
p
o
n
lin
e
class
i
f
ier
s
t
h
at
i
n
v
o
lv
e
o
n
l
y
a
s
m
al
l
an
d
f
i
x
ed
n
u
m
b
er
o
f
f
ea
tu
r
es.
R
e
s
u
lts
s
h
o
w
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
ar
e
f
air
l
y
ef
f
ec
ti
v
e
f
o
r
f
ea
t
u
r
e
s
ele
ctio
n
tas
k
s
o
f
o
n
l
in
e
ap
p
licati
o
n
s
,
an
d
s
ig
n
i
f
ica
n
tl
y
m
o
r
e
e
f
f
icie
n
t
a
n
d
s
ca
lab
le
th
an
s
o
m
e
s
tate
-
of
-
t
h
e
-
ar
t b
atc
h
f
ea
t
u
r
e
s
elec
t
io
n
tech
n
iq
u
e
[
2
8
]
.
6.
CO
NCLU
SI
O
N
Sp
litt
in
g
cr
iter
ia
o
f
d
ec
is
io
n
tr
ee
s
ar
e
ch
o
s
e
n
b
ased
o
n
s
o
m
e
q
u
a
lit
y
m
ea
s
u
r
es,
w
h
ic
h
r
eq
u
ir
es
h
an
d
li
n
g
th
e
e
n
tire
d
ata
s
e
t
o
f
ea
ch
ex
p
a
n
d
in
g
n
o
d
es.
T
h
is
m
ak
e
s
it
d
i
f
f
icu
l
t
f
o
r
d
ec
is
io
n
tr
ee
s
to
b
e
u
s
ed
in
b
ig
d
ata
ap
p
licatio
n
s
.
SVM
s
h
o
w
s
v
er
y
g
o
o
d
p
er
f
o
r
m
an
ce
to
d
ata
s
ets
in
a
m
o
d
er
ate
s
i
ze
.
I
t
h
as
in
h
er
en
t
li
m
ita
tio
n
s
to
b
ig
d
ata
ap
p
licat
io
n
s
.
Dee
p
lear
n
i
n
g
is
s
u
ited
t
o
ad
d
r
ess
is
s
u
e
s
r
elate
d
to
v
o
l
u
m
e
an
d
v
ar
iet
y
o
f
b
ig
d
ata.
Ho
w
ev
er
,
it
h
a
s
s
o
m
e
r
estrictio
n
s
i
n
b
ig
d
ata
b
ec
au
s
e
i
t
r
eq
u
ir
es
m
u
c
h
tr
ai
n
in
g
t
i
m
e.
P
L
ANE
T
ca
n
d
ea
l
w
it
h
b
ig
v
o
lu
m
e
o
f
d
ata,
b
u
t is n
o
t a
p
p
licab
le
to
d
ata
w
i
th
ca
teg
o
r
ical
a
ttrib
u
tes.
Ma
ch
i
n
e
lear
n
i
n
g
ap
p
licatio
n
s
in
b
ig
d
ata
h
as
m
et
ch
a
llen
g
e
s
s
u
c
h
as
m
e
m
o
r
y
co
n
s
tr
ai
n
t,
n
o
s
u
p
p
o
r
t
(
in
iter
atio
n
s
)
f
r
o
m
Ma
p
R
ed
u
ce
,
d
if
f
icu
l
t
y
in
d
ea
lin
g
w
ith
b
ig
d
ata
d
u
e
to
V
s
(
s
u
ch
as
h
i
g
h
v
elo
cit
y
,
v
o
l
u
m
e,
an
d
v
ar
iet
y
,
etc.
)
,
an
d
lear
n
i
n
g
tr
a
i
n
i
n
g
li
m
ited
to
a
ce
r
tain
n
u
m
b
er
o
f
clas
s
t
y
p
e
s
o
r
a
p
ar
ticu
lar
lab
eled
d
atasets
,
etc.
So
m
e
tech
n
o
lo
g
y
p
r
o
g
r
es
s
h
a
s
b
ee
n
m
ad
e
s
u
c
h
as
f
ac
eted
lear
n
in
g
f
o
r
h
ier
ar
ch
ical
d
ata
s
tr
u
ctu
r
e
,
m
u
lti
-
tas
k
lear
n
i
n
g
in
in
p
ar
allel,
m
u
lti
-
d
o
m
a
in
/
cr
o
s
s
-
d
o
m
ai
n
r
ep
r
esen
tatio
n
-
lear
n
i
n
g
,
s
tr
ea
m
i
n
g
d
ata
p
r
o
ce
s
s
in
g
,
h
ig
h
-
d
i
m
en
s
io
n
al
d
ata
p
r
o
ce
s
s
in
g
,
an
d
o
n
li
n
e
f
ea
tu
r
e
s
elec
tio
n
,
etc.
T
h
ese
ar
ea
s
an
d
th
e
ab
o
v
e
ch
alle
n
g
e
s
ab
o
u
t
m
ac
h
i
n
e
lear
n
in
g
i
n
b
ig
d
ata
also
ca
n
b
e
f
u
r
th
er
r
esear
ch
to
p
ics.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
Ma
ch
in
e
Lea
r
n
in
g
in
B
ig
Da
t
a
(
Lid
o
n
g
W
a
n
g
)
123
RE
F
E
R
E
NC
E
S
[1
]
C.
L
.
P
.
Ch
e
n
,
C.
Y.
Zh
a
n
g
,
“
Da
ta
-
in
ten
siv
e
a
p
p
li
c
a
ti
o
n
s,
c
h
a
ll
e
n
g
e
s,
tec
h
n
iq
u
e
s an
d
tec
h
n
o
l
o
g
ies
:
A
su
rv
e
y
o
n
Big
Da
ta
”
,
In
fo
rm
a
ti
o
n
S
c
ien
c
e
s
,
V
o
l
.
2
7
5
,
N
o
.
1
0
,
p
p
.
3
1
4
-
3
4
7
,
A
u
g
u
st 2
0
1
4
.
[2
]
K.
M
.
T
a
rw
a
n
i,
S
.
S
.
S
a
u
d
a
g
a
r,
H.
D.
M
isa
lk
a
r,
“
M
a
c
h
in
e
L
e
a
rn
in
g
in
Big
Da
ta
A
n
a
l
y
ti
c
s
:
A
n
Ov
e
r
v
ie
w
”
,
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
Res
e
a
rc
h
in
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
S
o
ft
wa
re
En
g
i
n
e
e
rin
g
,
Vo
l.
5
,
N
o
.
4
,
p
p
.
270
-
2
7
4
,
A
p
ril
2
0
1
5
.
[3
]
h
tt
p
s:/
/en
.
w
ik
ip
e
d
ia.o
rg
/w
ik
i/
M
a
c
h
in
e
_
lea
rn
in
g
[4
]
U.
Ja
s
w
a
n
t
a
n
d
P
.
N.
Ku
m
a
r,
“
Big
Da
ta
A
n
a
l
y
ti
c
s:
A
S
u
p
e
rv
ise
d
A
p
p
ro
a
c
h
f
o
r
S
e
n
ti
m
e
n
t
Clas
sif
ica
ti
o
n
Us
in
g
M
a
h
o
u
t:
A
n
Ill
u
stra
ti
o
n
”
,
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Ap
p
li
e
d
En
g
in
e
e
rin
g
Res
e
a
rc
h
,
Vo
l.
10
,
No
.
5
,
p
p
.
1
3
4
4
7
-
1
3
4
5
7
,
2
0
1
5
.
[5
]
Y.
De
m
c
h
e
n
k
o
,
P
.
G
ro
ss
o
,
C.
De
L
a
a
t
,
P
.
M
e
m
b
re
y
,
“
Ad
d
re
ss
in
g
Bi
g
D
a
ta
Iss
u
e
s
i
n
S
c
ien
ti
fi
c
Da
ta
In
fra
stru
c
t
u
re
”
,
2
0
1
3
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
Co
ll
a
b
o
ra
ti
o
n
T
e
c
h
n
o
l
o
g
ies
a
n
d
S
y
ste
m
s
(CT
S
)
,
S
a
n
Die
g
o
,
CA
,
USA
,
p
p
.
4
8
-
55
,
M
a
y
2
0
1
3
.
[6
]
D.
E.
O'
Lea
r
y
,
“
'
Bi
g
Da
ta
'
,
th
e
'
I
n
tern
e
t
o
f
T
h
in
g
s'
a
n
d
th
e
'
In
tern
e
t
o
f
S
ig
n
s'
”
,
In
telli
g
e
n
t
S
y
ste
ms
in
Acc
o
u
n
ti
n
g
,
Fi
n
a
n
c
e
a
n
d
M
a
n
a
g
e
me
n
t
,
V
o
l.
2
0
,
p
p
.
5
3
-
65
,
2
0
1
3
.
[7
]
H.
V
.
Ja
g
a
d
ish
,
A
.
L
a
b
rin
id
is
,
Y.
P
a
p
a
k
o
n
sta
n
ti
n
o
u
,
e
t
a
l
.
,
“
Big
Da
ta
a
n
d
Its
T
e
c
h
n
ica
l
Ch
a
ll
e
n
g
e
s
”
,
Co
mm
u
n
ica
ti
o
n
s o
f
th
e
ACM
,
Vo
l
.
57
,
No
.
7
,
p
p
.
86
-
94
,
2
0
1
4
.
[8
]
A
.
Zas
la
v
sk
y
,
C.
P
e
re
ra
a
n
d
D.
G
e
o
rg
a
k
o
p
o
u
l
o
s,
“
S
e
n
si
n
g
a
s
a
S
e
rv
ice
a
n
d
Bi
g
Da
t
a
”,
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
A
d
v
a
n
c
e
s in
Clo
u
d
C
o
m
p
u
ti
n
g
(
A
CC)
,
Ba
n
g
a
lo
re
,
In
d
ia,
p
p
.
1
-
8
,
Ju
ly
2
0
1
2
.
[9
]
M
.
T
u
rk
,
“
A
c
h
a
rt
o
f
th
e
b
ig
d
a
ta
e
c
o
s
y
ste
m
”
,
tak
e
2
,
2
0
1
2
.
[1
0
]
R.
Zaf
a
ra
n
i,
M
.
A
.
A
b
b
a
si,
H.
L
iu
.
“
S
o
c
ial
M
e
d
ia
M
i
n
in
g
:
A
n
In
tro
d
u
c
ti
o
n
”
,
Ca
m
b
rid
g
e
Un
iv
e
rsity
P
re
ss
,
UK
,
2
0
1
4
.
[1
1
]
J.
De
a
n
,
“
Big
Da
ta,
D
a
ta
M
in
in
g
,
a
n
d
M
a
c
h
i
n
e
L
e
a
rn
in
g
:
V
a
lu
e
Cre
a
ti
o
n
f
o
r
Bu
sin
e
ss
L
e
a
d
e
rs
a
n
d
P
ra
c
ti
ti
o
n
e
rs
”
,
Jo
h
n
W
il
e
y
&
S
o
n
s,
I
n
c
.
,
2
0
1
4
.
[1
2
]
K.
M
.
L
e
e
,
“
G
rid
-
b
a
se
d
S
in
g
le
P
a
ss
Clas
sif
ica
ti
o
n
f
o
r
M
ix
e
d
Big
Da
ta
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
p
p
li
e
d
En
g
i
n
e
e
rin
g
Res
e
a
rc
h
,
V
o
l
.
9
,
No
.
2
1
,
p
p
.
8
7
3
7
-
8
7
4
6
,
2
0
1
4
.
[1
3
]
M
.
M
.
Na
jafa
b
a
d
i,
F
.
V
il
la
n
u
stre
,
T
.
M
Kh
o
sh
g
o
f
taa
r,
N.
S
e
li
y
a
,
R.
W
a
ld
a
n
d
E.
M
u
h
a
re
m
a
g
i
c
,
“
De
e
p
lea
rn
in
g
a
p
p
li
c
a
ti
o
n
s an
d
c
h
a
ll
e
n
g
e
s in
b
ig
d
a
ta an
a
ly
ti
c
s
”
,
J
o
u
rn
a
l
o
f
Bi
g
D
a
ta
,
Vo
l.
2
,
No
.
1
,
2
0
1
5
.
[1
4
]
J.
W
.
Wan
g
,
Y.
T
a
n
g
,
M
.
Ng
u
y
e
n
,
I.
A
lt
in
tas
,
“
A
S
c
a
l
a
b
le Da
t
a
S
c
ien
c
e
W
o
rk
fl
o
w
Ap
p
ro
a
c
h
f
o
r B
i
g
Da
ta
B
a
y
e
sia
n
Ne
two
rk
L
e
a
rn
in
g
”
,
BDC
'
1
4
P
r
o
c
e
e
d
in
g
s
o
f
th
e
2
0
1
4
IEE
E
/A
CM
In
tern
a
ti
o
n
a
l
S
y
m
p
o
siu
m
o
n
B
ig
Da
ta
Co
m
p
u
ti
n
g
,
IE
EE
Co
m
p
u
ter
S
o
c
iety
,
W
a
sh
in
g
to
n
,
DC,
USA
,
p
p
.
16
-
25
,
2
0
1
4
.
[1
5
]
L
.
L
i,
“
Ex
p
e
ri
m
e
n
tal
Co
m
p
a
riso
n
s o
f
M
u
lt
i
-
c
las
s Clas
sif
iers
”
,
In
fo
rm
a
ti
c
a
,
Vo
l.
3
9
,
p
p
.
7
1
-
85
,
2
0
1
5
.
[1
6
]
S
.
B.
Ko
tsia
n
ti
s,
“
S
u
p
e
rv
ise
d
M
a
c
h
in
e
L
e
a
rn
in
g
:
A
Re
v
ie
w
o
f
Cla
ss
if
ic
a
ti
o
n
T
e
c
h
n
iq
u
e
s
”
,
In
fo
rm
a
t
ica
,
V
o
l.
3
1
,
p
p
.
249
-
2
6
8
,
2
0
0
7
.
[1
7
]
N.
Ka
ra
c
a
p
il
id
is,
M
.
T
z
a
g
a
ra
k
is
a
n
d
S
.
C
h
rist
o
d
o
u
l
o
u
,
“
On
a
m
e
a
n
in
g
f
u
l
e
x
p
lo
it
a
ti
o
n
o
f
m
a
c
h
in
e
a
n
d
h
u
m
a
n
re
a
so
n
in
g
to
tac
k
le d
a
ta
-
in
ten
siv
e
d
e
c
isio
n
m
a
k
in
g
”
,
In
telli
g
e
n
t
De
c
isio
n
T
e
c
h
n
o
l
o
g
ies
,
Vo
l.
7
,
p
p
.
2
2
5
–
2
3
6
,
2
0
1
3
.
[1
8
]
H.
Qia
n
,
“
P
iv
o
talR:
A
P
a
c
k
a
g
e
fo
r
M
a
c
h
in
e
L
e
a
rn
in
g
o
n
Big
Da
ta
”
,
T
h
e
R
J
o
u
r
n
a
l
,
Vo
l.
6
,
N
o
.
1
,
p
p
.
5
7
-
67
,
Ju
n
e
2
0
1
4
.
[1
9
]
A
.
Na
srid
in
o
v
,
“
Co
m
b
in
in
g
U
n
su
p
e
rv
ise
d
a
n
d
S
u
p
e
rv
ise
d
M
a
c
h
in
e
L
e
a
rn
in
g
to
A
n
a
ly
z
e
Crim
e
Da
ta
”
,
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
A
p
p
li
e
d
En
g
i
n
e
e
rin
g
Res
e
a
rc
h
,
V
o
l
.
9
,
No
.
2
3
,
p
p
.
1
8
6
6
3
-
1
8
6
6
9
,
2
0
1
4
.
[2
0
]
S.
Ka
n
a
g
a
v
a
ll
i,
S
.
V
a
ish
a
li
,
J.
L
.
Je
b
a
,
“
A
n
a
l
y
sis
a
n
d
M
i
n
in
g
o
f
S
o
c
ial
Ne
tw
o
rk
Da
ta
F
o
r
S
o
c
iety
I
ss
u
e
s
B
y
Us
in
g
Big
Da
ta
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
p
p
li
e
d
E
n
g
i
n
e
e
rin
g
Res
e
a
rc
h
,
Vo
l.
10
,
No
.
4
,
p
p
.
1
0
4
9
7
-
1
0
5
0
6
,
2
0
1
5
.
[2
1
]
S
.
R.
S
u
k
u
m
a
r,
“
M
a
c
h
in
e
L
e
a
rn
i
n
g
i
n
t
h
e
Bi
g
Da
t
a
Era
:
Are
W
e
T
h
e
re
Y
e
t?
”,
A
CM
Kn
o
w
led
g
e
Disc
o
v
e
r
y
a
n
d
Da
ta
M
in
in
g
:
W
o
rk
sh
o
p
o
n
Da
t
a
S
c
ien
c
e
f
o
r
S
o
c
ial
G
o
o
d
,
Oa
k
Rid
g
e
Na
ti
o
n
a
l
L
a
b
o
ra
to
ry
,
p
p
.
1
-
5
,
De
c
e
m
b
e
r
2
0
1
4
.
[2
2
]
Y.
Bu
,
V
.
B
o
rk
a
r,
M
.
J.
Ca
re
y
,
J.
Ro
se
n
,
N.
P
o
ly
z
o
ti
s,
T
.
Co
n
d
ie,
M
.
W
e
ime
r,
R.
Ra
m
a
k
rish
n
a
n
,
“
S
c
a
li
n
g
Da
talo
g
f
o
r
M
a
c
h
in
e
L
e
a
rn
in
g
o
n
B
ig
Da
ta
”
,
M
a
rc
h
2
0
1
2
.
[2
3
]
S.
S
u
t
h
a
h
a
ra
n
,
“
Big
Da
ta
Clas
si
fica
ti
o
n
:
P
ro
b
lem
s
a
n
d
Ch
a
ll
e
n
g
e
s
in
Ne
t
w
o
rk
In
tru
sio
n
P
re
d
icti
o
n
w
it
h
M
a
c
h
in
e
L
e
a
rn
in
g
”
,
Per
fo
rm
a
n
c
e
Eva
lu
a
ti
o
n
Rev
iew
,
Vo
l.
41
,
No
.
4
,
p
p
.
7
0
-
73
,
M
a
rc
h
2
0
1
4
.
[2
4
]
S
.
Hid
o
,
S
.
T
o
k
u
i,
S
.
Od
a
,
“
Ju
b
a
tu
s:
A
n
Op
e
n
S
o
u
rc
e
P
latf
o
rm
f
o
r
Distrib
u
ted
On
li
n
e
M
a
c
h
in
e
L
e
a
rn
in
g
”
,
T
e
c
h
n
ica
l
Re
p
o
rt
o
f
th
e
Jo
in
t
Ju
b
a
tu
s
p
ro
jec
t
b
y
P
re
f
e
rre
d
In
f
r
a
stru
c
tu
re
In
c
.
,
a
n
d
NT
T
S
o
f
t
wa
re
In
n
o
v
a
ti
o
n
Ce
n
ter,
T
o
k
y
o
,
Ja
p
a
n
,
N
I
P
S
2
0
1
3
W
o
r
k
s
h
o
p
o
n
B
i
g
L
e
a
r
n
i
n
g
,
L
a
k
e
T
a
h
o
e
,
p
p
.
1
-
6
,
D
e
c
e
m
b
e
r
2
0
1
3
.
[2
5
]
Y.
L
o
w
,
J.
G
o
n
z
a
lez
,
A
.
K
y
ro
la,
D.
Bick
so
n
,
C.
G
u
e
strin
,
J.
M
.
H
e
ll
e
rste
in
,
“
Distrib
u
te
d
Gr
a
p
h
L
a
b
:
A
Fra
me
wo
rk
fo
r
M
a
c
h
i
n
e
L
e
a
rn
in
g
a
n
d
Da
t
a
M
in
in
g
i
n
t
h
e
Clo
u
d
”
,
T
h
e
3
8
t
h
In
ter
n
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
Ve
r
y
Larg
e
Da
t
a
Ba
se
s
,
En
d
o
w
m
e
n
t
,
Vo
l.
5
,
No
.
8
,
p
p
.
7
1
6
-
7
2
7
,
2
0
1
2
.
[2
6
]
A
.
B
a
ld
o
m
in
o
s,
E.
A
lb
a
c
e
t
e
,
Y.
S
a
e
z
a
n
d
P
.
Isa
si,
“
A
S
c
a
la
b
le
M
a
c
h
in
e
L
e
a
rn
i
n
g
On
li
n
e
S
e
rv
ice
fo
r
Bi
g
Da
ta
Rea
l
-
T
ime
An
a
lys
is
”
,
2
0
1
4
IEE
E
S
y
m
p
o
siu
m
o
n
Co
m
p
u
tatio
n
a
l
In
telli
g
e
n
c
e
in
Big
Da
ta
(CIBD):
p
r
o
c
e
e
d
in
g
s
,
pp.
1
-
8
,
2
0
1
4
.
[2
7
]
A
.
T
.
V
u
,
G
.
De
F
.
M
o
ra
les
,
J.
G
a
m
a
,
A
.
Bif
e
t,
“
Distrib
u
ted
A
d
a
p
ti
v
e
M
o
d
e
l
Ru
les
fo
r
M
in
in
g
Bi
g
Da
ta
S
tre
a
ms
”
,
2
0
1
4
IE
EE
I
n
tern
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
B
ig
Da
ta
,
W
a
sh
in
g
to
n
,
D
C,
p
p
.
3
4
5
-
3
5
3
,
Oc
t
o
b
e
r
2
0
1
4
.
[2
8
]
S.
HO
I,
J.
W
a
n
g
,
P
.
Zh
a
o
,
a
n
d
R.
Jin
,
“
On
li
n
e
Fea
t
u
re
S
e
lec
ti
o
n
f
o
r
M
in
in
g
Bi
g
Da
ta
”
,
Big
M
i
n
e
'
1
2
P
r
o
c
e
e
d
in
g
s o
f
th
e
1
st
I
n
tern
a
ti
o
n
a
l
W
o
rk
sh
o
p
o
n
Big
Da
ta,
S
trea
m
s
a
n
d
He
tero
g
e
n
e
o
u
s
S
o
u
rc
e
M
i
n
in
g
:
A
lg
o
ri
th
m
s,
S
y
ste
m
s
,
P
r
o
g
ra
m
m
in
g
M
o
d
e
ls
a
n
d
A
p
p
li
c
a
ti
o
n
s
,
A
CM
Ne
w
Yo
rk
,
NY
,
USA
,
p
p
.
9
3
-
1
0
0
, 2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.