►
Description
FINN.no is the leading classifieds website in Norway and the country's busiest website. At FINN.no fraud is a real, ongoing, constantly changing problem that continuously requires a lot of resources. Plus, fraud is an area of development that's not frequently discussed. Mick will go through various challenges, ideas, and thoughts about where and why Cassandra is the best choice when working with machine learning.
A
A
A
tremendous
supporter
of
the
Cassandra
community
throughout
Europe
he
is
suspected
in
new
charts
and
he
at
the
London
Cassandra
summit
last
year,
so
met.
We
really
appreciate
your
support
of
the
Cassandra
community,
very
excited
to
have
you
with
us
today
and
if
you
would
like
to
ask
make
a
question,
we
will
hold
all
the
questions
until
the
end
of
the
webinar.
A
B
A
B
You
Christian,
so
if
in
n.o
is
nowhere
busiest
website
and
it
is
a
classifieds
website,
it
is
the
predominant
classifieds
website
of
a
very
broad
range
of
octopods.
So
here
we're
talking
real
estate,
jobs,
bits
and
pieces
like
trading,
post
or
ebay.
Our
foes
motorcycles
travel
services
like
cheapest
flights,
cheapest
hotel
and
personal
and
professional
services.
The.
A
B
A
B
B
B
A
B
What
we
found
afterwards
was
that
cassandra
is
a
fantastic
database
which
can
be
used
for
a
lot
of
other,
simpler
use
cases,
and
so
today,
when
people
come
to
us
and
they
say,
hey
we're
riding
to
new
code
or
we're
redesigning
some
old
code.
Do
you
think
that
cassandra
is
the
right
choice?
The
questions
that
we
ask
pass
them
along
resigns?
B
Do
you
have
a
high
or
fast
write,
throughput
requirements?
If
you
do
in
Cassandra's
rapturously,
can
you
see
yourself
benefiting
from
a
schema
free
design
if
you're
doing
this
to
elder
thermal
and
you're,
finding
that
you're
using
join
or
if
you
see
that
you
have
sparse
data
or
that
you
have
growth
in
columns
as
much
as
you
have
growth
and
rose
again,
Cassandra's.
A
B
A
B
B
B
A
B
And
we
also
don't
run
commodity
machines,
so
our
hosting
provider
only
gives
us
quite
beeping
machines,
so
I
Cassandra
cluster
is
of
just
six
notes,
spread
over
to
physical
data
centers
for
all
those
nodes,
a
paid
residue
like
I
mentioned.
The
first
use
case
worked
with
both
Cassandra
Hannah
do
and
we
had
two
extra
notes
which
don't
have
to
do
any
lemons.
That's
because
our
Cassandra
cluster
needs
24
so
is
serving
request
with
the
latency
low
enough
web
applications,
and
we
tackle
that
by
creating
for
virtual
data
centers,
so
that
each
node.
B
A
B
A
B
B
B
A
B
A
B
B
A
B
Inside
the
fraud
control
system,
we
have
like
an
expert
rules
sister,
so
we
have
a
set
of
rules
which
are
applied
against
each
add.
The
scores
from
those
rules
are
weighted
together
and,
depending
on
the
final
score,
the
theme
is
made.
A
low
score
will
result
in
and
being
automatically
publish.
The
middle
range
scroll
will
result
in
the
ad
going
to
our
our
soft
manual
fraud.
Roll
and
a
high
school
means
that
it
goes
to
our
in-house
fraud
and
forms.
B
I'm
a
complicated
diagram
for
this
process.
You
can
see
that
when
I'm
added
ended
comes
into
the
fraud
control
system
and
it's
given
a
9v,
it's
based
on
the
ad
ID
and
the
timestamp
is
change.
The
rules
are
applied
and
recorded
against
that
that
ID
decision
is
made
published
to
end
up
as
publish
invalid
or
rejected,
and
so
the
published
add
if
it
edited
by
the
owner.
It
goes
back
in
so
then
the
second
time
when
it
gets
scored,
it
gets
scored
under
a
new
ID,
because
the
Prime
staff
changes.
B
A
B
B
A
B
The
clothes
and
is
dealt
with
here,
this
data,
with
interested
in
looking
at
how
the
experts
rules
system
is
operating
over
periods
of
time,
the
first
to
lex
payment.
We
do
is
a
given
day
get
me
all
of
the
ads
which
have
been
calculated
and
all
the
rules
and
their
scores
from
that
list
of
a
tidy
get.
A
A
B
A
B
A
A
A
B
B
A
B
B
A
B
A
B
A
B
B
B
A
A
B
B
A
B
A
A
B
A
A
A
B
B
A
A
B
A
A
A
B
A
A
B
Be
testing
could
get
you
to
interrupt
a
loop
just
like
a
salt
chasing
its
own
tail,
it's
better
to
take
a
more
complicated
to
look
at
a
be
testing
or
multivariate
testing.
You
have
different
training
sets
of
data.
You
have
different
algorithms
sore
arms
tests
continue
when
you
start
to
see
the
need
for
that.
You
start
to
see
the
need
for
a
better
machine
learning
platform,
and
this
moves
on
to
the
next
point
in
that,
when.
B
B
B
B
A
B
Live
in
oslo
or
you
do
live
in
oslo,
please
dig
in
contact
with
us.
The
point
of
the
presentation
was
to
give
you
a
run-through
of
cassandra
and
some
data
models
that
we
have
in
production.
Do
some
experience
about
Cassandra
setup,
but
also
to
show
that
while
machine
learning
and
series
can
be
simple,
when
you
start
bringing
into
the
picture
ground.
A
B
A
A
B
A
Thank
you
very
much,
so
it's
quite
it.
You
can
ask
questions,
but
it
sounds
like
Mick
might
have
to
kill
you
if
you
do
so
a
couple
of
upcoming
events,
while
we
are
receiving
any
questions
and
make
small
crowd
on
hand
today,
so
we
may
not
get
any
questions,
but
we'll
hang
on
here
for
a
few
minutes.
The
next
webinar
July
can
buckle
up
the
future
of
the
distributed
databases.
Here,
datastax
is
very
own.
Robyn,
Schumacher
and
martin
van
ryswyk
be
presenting
on
that
and
some
exciting
developments
will
be
announced
there.
A
And
then,
if
you
want
to
continue
your
cassandra
education,
we
have
our
free
online,
co-founder
training
and
then
do
not
miss
bar
in
that
states.
Do
not
miss
the
Cassandra
summit
2014,
which
is
in
San
Francisco
on
sep,
tember,
tenth
and
eleventh,
and
we
have
some
certified
training
on
day
one
and
then
on
day,
two.
We
have
a
fantastic
agenda
that
we
will
be
announcing
actually
later
this
week,
so
make
sure
to
check
out
the
microsite
on
dates.
Tax
com
make.
Are
you
able
to
make
it
over
this
year?.
A
Yeah
saw
that
come
through,
and-
and
so
if
you
are
in
Europe
no
worries
at
all,
the
Cassandra
summit
will
come
to
Europe
in
the
depth
of
winter,
because
we
because
why?
Why
not
so
do
we
have
any
questions
here?
Let's
have
a
look
so
market
just
one
question
right
now
and
that
is
from
market
mall,
dashi
Machel
and
he
is
asking
nick.
Did
you
consider
using
MongoDB?
So
maybe
maybe
you
can
address
that
directly,
but
and
maybe
talk
about
some
of
your
criteria
for
selecting
Cassandra.
A
A
B
A
A
A
B
A
Yep.
Can
you
do
that
page
and
then
we
will.
We
will
wrap
up
there
today
make
and
we
do
love
to
have
you
back
to
you
know
present.
This
may
be
closer
to
you,
the
Cassandra
summit,
and
you
know
thank
you
again
for
being
such
a
great
supporter
of
the
co-founder
community.
We
really
really
appreciate
it.
Thank.