http://arxiv.org/abs/1402.2308
Presented at 23rd International World Wide Web Conference, Seoul, Korea, April, 2014
With public information becoming widely accessible and shared on today's web, greater insights are possible into crowd actions by citizens and non-state actors such as large protests and cyber activism. We present efforts to predict the occurrence, specific timeframe, and location of such actions before they occur based on public data collected from over 300,000 open content web sources in 7 languages, from all over the world, ranging from mainstream news to government publications to blogs and social media. Using natural language processing, event information is extracted from content such as type of event, what entities are involved and in what role, sentiment and tone, and the occurrence time range of the event discussed. Statements made on Twitter about a future date from the time of posting prove particularly indicative. We consider in particular the case of the 2013 Egyptian coup d'etat. The study validates and quantifies the common intuition that data on social media (beyond mainstream news sources) are able to predict major events.
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Predicting Crowd Behavior with Big Public Data
1. Predic'ng
Crowd
Behavior
with
Big
Public
Data
Nathan
Kallus
Massachuse.s
Ins0tute
of
Technology
As
seen
on:
April
8,
2014
23rd
Interna0onal
World
Wide
Web
Conference
Seoul,
Korea
2. How
did
crowds
used
to
come
together
and
how
did
we
hear
about
it
in
the
analog
era?
3. How
did
crowds
used
to
come
together
and
how
did
we
hear
about
it
in
the
analog
era?
4. How
did
crowds
used
to
come
together
and
how
did
we
hear
about
it
in
the
analog
era?
5. How
did
crowds
used
to
come
together
and
how
did
we
hear
about
it
in
the
analog
era?
6. How
did
crowds
used
to
come
together
and
how
did
we
hear
about
it
in
the
analog
era?
9. “Since the revolt in Syria, the security
situation in Lebanon has deteriorated.”
The
Growth
of
Data
10. “Since the revolt in Syria, the security
situation in Lebanon has deteriorated.”
The
Growth
of
Data
11. “Since the revolt in Syria, the security
situation in Lebanon has deteriorated.”
The
Growth
of
Data
12. “Since the revolt in Syria, the security
situation in Lebanon has deteriorated.”
The
Growth
of
Data
13. Is
this
data
predic0ve?
• Sunday
6/9/13:
One
protester
dead
in
a
violent
Beirut
protest
against
Hezbollah's
interference
in
Syria
• 1
day
before
in
the
news:
“Lebanese
fac0on
organizes
two
demonstra0ons
tomorrow
rejec0ng
the
par0cipa0on
of
Hezbollah
in
the
figh0ng
in
Syria”
(translated
from
Arabic)
• 4
days
before
on
Twi.er:
“Say
no
to
#WarCrimes
and
demonstrate
against
#Hezbollah
figh0ng
in
#Qusayr
on
June
9
at
12
PM
in
Downtown
#Beirut”
• General
sense
of
violence
through
news
fragments:
6/6:
“Fatwa
Calls
For
Suicide
A.acks
Against
Hezbollah”
(TheBlaze.com)
6/4:
“Since
the
revolt
in
Syria,
the
security
situa0on
in
Lebanon
has
deteriorated”
(Al
Bawaba)
5/23:
“The
revolt
in
Syria
has
exacerbated
tensions
in
Lebanon,
which
...
remains
deeply
divided”
(Huff
Post)
14. The
signal
is
there…
Reports
of
protest
in
Lebanon
by
publish
day:
We
just
need
the
data…
Mainstream news
Forward-looking twitter
9ê15 10ê1 10ê15 11ê1 11ê15
Day
50
100
Mentions
15. Data catered by Recorded Future
Con0nually
scans
300,000+
sources
in
7
languages.
News,
blogs,
social
media,
govt
publica0ons…
18. Use
this
to
quan0fy
the
signals
All Mainstream Twitter
Afghanistan 60918 13979 27655
Bahrain 246136 32873 177310
Egypt 944998 246882 397105
France 172508 22648 111702
Greece 122416 18037 70521
India 491475 56981 274027
Indonesia 34007 6870 17120
Iran 118704 26487 53962
Italy 65569 8977 43803
Jordan 35396 7991 19369
Lebanon 44153 9610 23394
Libya 162721 43093 69437
Nigeria 70635 7873 38700
Pakistan 289643 25982 213636
Saudi Arabia 39556 12452 13670
Sudan 28680 6733 13654
Syria 212815 63538 79577
Tunisia 99000 35218 27233
Yemen 70583 29140 16712
• 19
countries
• Events
published
1/1/2011–7/10/2013
• Millions
of
reports
of
protests!
• Train
(+
cross-‐val)
on
1/1/2011–3/5/2013
• Test
on
3/6/2013–7/10/2013
19. • Clustering
of
country
in
ques0on
(hierarchical
clustering
on
Kolmogorov
distance)
• Same-‐day
mainstream
reports
of
protests
over
past
10
days
• Level
of
violence
language
in
those
• Forward-‐looking
events
reported
on
Twi<er
about
days
in
ques0on
posted
over
past
10
days
• Forward-‐looking
events
reported
by
mainstream
sources
about
days
in
ques0on
posted
over
past
10
days
Features
for
Random
Forest
To
determine
whether
a
future
0me
will
have
significant
protests,
base
our
predic0on
on…
20. Results
0% 25% 50% 75% 100%
FPR
25%
50%
75%
100%
TPR
0% 25% 50% 75% 100%
FPR
25%
50%
75%
100%
TPR
Locale-‐rela0ve
scale
of
significance
Global
absolute
scale
of
significance
(23%
of
posi0ve
training
instances
in
Egypt)
21. Case
Study:
2013
Egyp0an
Coup
• 6/30/13:
Anniversary
of
Morsi’s
rule
– Protests
broadly
an0cipated
(even
Kerry
made
a
statement
asking
for
calm)
• 6/28-‐6/29/13:
Unan0cipated
warm-‐up
protests
• 7/3/13
and
onward:
Morsi
removed
from
power,
nonstop
protes0ng
and
violence
persists
for
weeks
to
come
24. From
a
March
26
email
“Right
now
I
think
we
can
likely
expect
something
to
happen
in
Egypt
on
Friday.
There've
been
protests
so
far
…
but
these
have
been
at
Egypt's
"baseline"
levels
so
far.
With
tensions
rising
and
calls
on
twi.er
from
the
pro-‐
Brotherhood
side
as
well
as
counter
calls
for
celebratory
demonstra0ons
from
the
pro-‐Sisi
side
ci0ng
Tamarod
…
This
will
bubble
up
further
throughout
today
and
likely
erupt
on
Friday.”
43
HOURS
LATER:
(The
previous
big
coverage,
a
week
before,
about
unrest
in
Egypt
included
a
single
death.
I.e.,
this
is
significant.)
A
more
recent
predic0on
25. From
a
March
26
email
“Right
now
I
think
we
can
likely
expect
something
to
happen
in
Egypt
on
Friday.
There've
been
protests
so
far
…
but
these
have
been
at
Egypt's
"baseline"
levels
so
far.
With
tensions
rising
and
calls
on
twi.er
from
the
pro-‐
Brotherhood
side
as
well
as
counter
calls
for
celebratory
demonstra0ons
from
the
pro-‐Sisi
side
ci0ng
Tamarod
…
This
will
bubble
up
further
throughout
today
and
likely
erupt
on
Friday.”
Facts
men0oned
in
AP
release
at
a
later
0me
• Track
those
social
media
trends
that
will
lead
to
a
significant
event
to
be.er
understand
what’s
going
on
and
what’s
going
to
go
on
and
why
A
more
recent
predic0on
27. From
an
April
1
email
“…
upcoming
protest
in
Bahrain.
The
tweets
behind
that
predic0on
reveal
it
to
be
an0-‐government
demonstra0ons
planned
to
start
ahead
of
the
upcoming
F1
Grand
Prix.”
56
HOURS
LATER:
A
more
recent
predic0on
28. More
Results:
Predic0ng
Hack0vist
Cyber
A.acks
Targets BAC Perpetrators BAC
Israel 68.9% Anonymous 70.3%
Germany 65.4% AnonGhost 70.8%
South Korea 63.1% LulzSec 60.6%
United Kingdom 65.5% Guccifer 66.7%
By
targets
or
perpetrator.