Weka Tutorial 34: Generating Stratified Folds (Data Preprocessing) Video Lecture | Weka Tutorial - Data & Analytics

hi this tutorial I'm going to show you
how you can remove stratified folds from
your data set in one of my tutorials
before I have shown you that how you can
remove folds from your data set but
those folds were not stratified so
stratification means that the
distribution of the class labels in your
data set they should be maintained in
each of your folds you are going to
create so here for example the years the
number of yes instances in my data set
here is about 372 and number of no
instances are about 6000 and 500 so you
can see you can easily count the ratios
of years to know there and if I
calculate them here right now you can
see that I have six thousand four
hundred and forty nine yeses sorry nose
and 372 yeses then the ratio here is
about 17 so the a ratio of no to es is
17 so when you're going to make when
you're going to set up a stratified
k-fold cross-validation each of your que
polls should have instances where the
ratio is kept just like this around 17
16.9 can be 17.5 that is also valid
ratio of the multi years because it's
around 17 point which is not exactly
seventeen point three three so that's
that's what what we call a
stratification so how you can do that
how are you going to remove the falls
from your data set in a stratification
manner so in order to do that we go to
wake up filters and supervised filters
instance filters and you can see that
there is an option called stratified
remove folks I choose that and if I
click a left click on on to edit the
properties you can see that here are
four options
fold-in vertices
election number of voles and seed so we
are not going to talk about inverse
selection and seed here because you can
always take a look at those options from
by clicking the more button here I'm not
going to talk about them so fall is the
full number you're going to create the
default is one that means that in these
out of these ten foals you're going to
create the first of your folds and if I
click OK and click apply you can see
that everything has changed so this is
my first fold of ten folds and this
first fold contains about six hundred
and eighty-three instances you can see
that if you multiply that by ten is
going to be around the number of
instances in your data set if I go to
class then you can see the number of ESS
now went down to 38 and number of no
went down to 645 so we have our
calculator here and we're just trying to
get the ratio again to see if it is
working really well or not so 645 knows
divided by 38 ESS it's sixteen point
nine seven or seventeen the ratio is
about seventeen where the ratio of no
two years in my entire data set was
about seventeen point three so that's
that's okay so that's how it works that
these are stratified Falls you're going
to create and if you want to save it
then you can always click on Save button
and you can rename that our file to
anything like full number one for number
one and so on so now I'm not going to
save that to produce the second volt
you're going to left click on these and
this filter and you're going to put two
there so that's your for number two
you're going to create the full number
two there but two before doing that I
have to undo that and do this and you
have to go back to your original data
set containing 6821 instances you click
left click and you select to that's
you're going to create the full number
- click ok and press apply you see that
now the number of instances went down to
682 and also the class distribution is
almost the same as the class
distribution in your original dataset so
I hope that helps that I'll be back with
more tutorials very soon thanks a lot