►
From YouTube: ML.NET Tutorial: How to Scale Numerical Data
Description
This video will show how to scale numerical data using ML.NET to help you get better performance out of your machine learning algorithms.
Code - https://github.com/jwood803/MLNetExamples/blob/master/MLNetExamples/ScaleData/Program.cs
ML.NET Transforms playlist - https://www.youtube.com/playlist?list=PLl_upHIj19ZyJf-ItL5Mey9mniPiTKtvv
ML.NET Playlist - https://www.youtube.com/watch?v=8gVhJKszzzI&list=PLl_upHIj19Zy3o09oICOutbNfXj332czx
Contact:
Twitter: https://twitter.com/JWood/
Blog: https://jonwood.co/
Gear used (affiliate links):
Mic - https://amzn.to/2YEXtxI
Mouse - https://amzn.to/2ZtASoQ
A
Hey
everyone
in
this
video
I
want
to
talk
about
how
you
can
scale
your
numerical
data
within
on
that
net.
First
of
all,
let's
go
over
why
you
want
to
scale
your
data
in
the
first
place,
so
I'm
gonna,
look
at
my
data
and
I'm
using
the
housing
data
set
here
and
just
look
at
this
row.
I
have
highlighted
and
look
at
some
of
the
numbers
here,
so
we
have
a
three
thousand
four
total
rooms.
A
But
if
we
look
at
the
median
income,
it's
just
three,
so
this
might
be
in
terms
of
thousands
or
ten
thousand
or
something
like
that.
But,
as
you
can
see,
three
is
thousand
times
less
than
three
thousand.
So
what
might
happen
with
the
machine
learning
algorithm
was
it
might
take
this
column
of
total
rooms
might
give
it
more
weight
than
the
other
columns,
especially
this
median
income
column.
A
Just
because
it's
that
much
bigger
and
what
the
reality
is
for
this
data
set
is
that
perhaps
the
median
income
might
have
a
greater
relationship
on
the
median
house
Belliard
that
we're
trying
to
predict.
But
it
won't
do
that
if
it
gives
it
more
weight
on
the
total
rooms
so
to
overcome
that
we
can
scale
our
data
and
when
we
scale
our
data,
we
can
change
all
the
numerical
values
between
0
and
1
that
way
it'll.
A
So
back
in
visual
studio
here
have
console
project
ready
and
already
have
emailed
Annette
installed
version,
1.3
point
1
and
I.
Have
the
data
set
and
I
have
the
input
schema
as
well
as
the
output
schema
class
was
already
done
so
to
get
started,
we're
going
to
create
a
context.
I'm
gonna
read
the
data
from
the
file
here.
A
And
then
we
get
a
collection
of
columns
kind
of
similar
to
get
in
the
features,
but
I
want
to
get
the
columns
that
we're
going
to
use
for
kind
of
the
input
for
our
scaling
here.
To
do
that,
we
do
data,
that's
schema
and
we'll
use
link
to
do
select
just
you
to
get
the
column
name
and
then
I
want
to
filter
these
out.
A
Or
if
it's
the
ocean
proximity,
because
I
don't
want
the
label
I'll
need
to
scale
that
and
I
don't
need
the
ocean
proximity,
because
it's
a
categorical
and
remember
the
label
and
ocean
proximity
doesn't
match
what
comes
in
here
and
matches.
What's
in
our
input,
schema
here
so
had
median
house
value
that
it
reads,
but
I'm
renaming
it
to
the
label
and
then
the
ocean
proximity
I
think
that
I'm
gonna
select
again.
A
This
time
for
each
of
these
items
and
get
a
new
input,
output,
column
pair
and
I
just
do
the
same
column
name
for
each
of
the
input
and
output
column
names
and
that
I'll
just
set
that
to
an
array
and
to
scale
our
data
would
do
the
context
transforms
and
they
would
be
normalize,
I'm
doing
normalize,
min
and
Max.
But
you
see
there
are
quite
a
few
scaling
methods
here,
I
just
put
in
those
columns
and
then.
A
So
now
that
I
have
my
scaled
data
now
to
kind
of
give
a
bit
of
a
look
at
it.
We'll
look
at
the
preview
and
actually
I'm,
just
gonna
run
a
breakpoint
here
and
see
what
this
looks
like
just
to
show
you
what
the
data
looks
like
once
it's
been
scaled
and
we
will
step
over
that.
Let's
look
at
the
road,
you
look
at
the
first
row,
so
you
can
see
we
have,
since
we
named
that
the
input
and
output
the
same
we're
gonna
get
double
hearing.
A
A
A
Still
I'll
put
in
the
features
now
send
in
that
this
text
column
that
we
created
before
and
then
I'm
going
to
use
the
regression
training
I'll
do
the
voice
on
regression
and
since
we
have
label
column
and
a
features
column,
we
can
just
leave
the
defaults
for
the
constructor
there
all
right.
So
we
have
our
pipeline
that
we
can
create
our
model
fitting
on
it
and
we're
not
just
going
to
do
data
we're
gonna.
Do
our
scale.
Data.
A
And
then
we
can
create
a
prediction
engine
passing
ammonal.
Now
we
can
create
a
prediction
from
the
prediction:
engine
I'll
just
paste
in
an
example
here,
just
random
kind
of
random
values,
and
then,
with
our
prediction,
I
can
write
out
what
we
get
there
put
it.
The
house
value
now
do
a
console.readline,
so
it
stays
up
all
right.
So
let's
run
this
and
see
what
we
get
all
right.
So
with
that
data
we
get
a
predicted
house
value
of
$259,000.
A
All
right,
I
think
I'll
end
things
there,
just
to
kind
of
show
you
how
you
can
use
the
transform
to
scale
your
numerical
data,
meaning
just
remembering
that
all
these
transform
the
numerical
data
between
0
&
1.
So
they
don't
have
that
orders
of
magnitude,
difference
between
one
column
and
another
column.
All
right
thanks.
Everyone
and
I'll
see
you
all
next
time.