Weka Tutorial 08 Numeric Transform (Data Preprocessing) Video Lecture -

hello in this way a tutorial this is the
eight-week a tutorial we are going to
see how we can use the numeric transform
filter in wake up so in order to do the
do this we are going to use the I
restored our file provided with the
wake-up data sets so in this I restored
our file you can see that we have five
features or attributes civil length
sepal width pedaling pedal weight and
class here you can see that sibling they
are considered as real numbers and we
can have some data here like five point
one four point nine and so on but this
is very realistic that when you for your
particular data set are extracting some
features from text so from images then
you can have real numbers that goes 5 to
7 or to 15 digits even after the decimal
points so if you have a large number of
precision software decimal points then
it has been proved that some of the
classification algorithms they give you
bizarre results and in order to avoid
them you have to transform those feature
values aida to its absolute values or to
ceilings or to the floors so that your
classification algorithms can really
read those feature values appropriately
and develops a better classifier for you
so in order to use the numeric transform
in a way that we can explore we are
loading for this tutorial the iris
dataset
so in this Irish data said if you choose
the florist feature for example Sybil
length you can see that this selected
attribute has the minimum value of four
point three maximum of seven point nine
and you also have some information on
the mean and standard deviation for this
particular feature so what if your
classification algorithm really doesn't
support Precision's after decimal points
that means in other words if you give
your classification algorithm some
positions of the decimal points then
it's it's not learning at all at all it
only works well with an integer not with
the real number so what you do that time
so in order to avoid that you can use
the waik enumerate transform filter and
you know that you do that you can choose
and this filter go to the wake up
filters unsupervised and attribute there
you can find somewhere in the middle a
filter called numeric transform so you
select them and when you are selecting
you will see that you have a default
setting of method name called apps
so this apps is the exact method that's
defined inside Java and so far from
these tutorials you already are assuming
that you already know that Vega is built
up from Java so in this case in to
transform those feature values
numerically it uses some of the Java
methods so abs or absolute method of
Java is the default setting for numeric
transform filter so what we are going to
do here is we are changing apps to flow
and you should know that Java has a math
method called floor that actually takes
the floor value of any particular
instance and as we are trying to
transform the future value of spam
sample length which is our first feature
in attribute in this's you define that
to one so if you want to change if you
want to transform the numeric values of
petal width then you should write it
down for here okay so as we are
interested for changing or transforming
the numeric values of several lengths or
our first feature for this tutorial
we're going to use attribute indices one
and you can see that the class named
here is Java dogleg map that's that's
the default Java class and we have used
the mathematical floor so we click OK
and we click apply so you can see that
the minimum and maximum they are now all
set to some floor values
this means that previously the minimum
was 4.3 because we were having some real
numbers in our data set but as we have
used or applied damaris form with a
floor java method at that time we are
not currently having any kind of real
number we are all having the integer
numbers or this particular attribute C
if you want to save this data set at
this stage and we are going to save that
on my desktop we save that and we're
going to double check if we have done
everything right
so we are going to open it with notepad
plus plus and you can see that the first
attribute value they are not containing
the decimal points anymore so in this
way if you have any classification
algorithm that does not support
precision numbers of the decimal point
so that does not support real numbers
but when you are extracting features
from your data set that those that are
actually real numbers at that time you
can use the numeric transform filter of
wake up and you can choose you can
choose absolute values you can choose
floors and you can choose ceilings in
order to transform the numeric features
I hope that helps thank you