Weka Tutorial 18: Classification 101 with Knowledge Flow Environment (Classification) Video Lecture | Weka Tutorial - Data & Analytics

hello youtube and welcome to the 18th
waker tutorial in this tutorial i will
start playing with the third application
in Vega GUI chooser
it's name is knowledge flow so far we
dealt with Explorer and we also have
seen some of the functionalities related
with experimenter what you have done
with Explorer is you can always analyze
your data and you can do some sanity
checks or send your tests before
indulging yourself into too much deep
into applying any particular classifier
on a particular data set and with the
experimenter we have seen that we can
have the facility to do more than what
we have done with Explorer the knowledge
flow is an application provided with
waked GUI and it's it's been fantastic
because with knowledge flow you can do
almost everything actually you can do
everything that you could do with the
Explorer application plus with knowledge
flow you can have the facility to
visualize what you are currently doing
and also you can you can save at the
settings the methods you're currently
following to find out the beautiful
results you were expecting with your
data set and your classifiers so this is
this is an enhanced tool you can say
from the Explorer that provides you the
facility of visualization of what you
are actually doing so I believe that
you'll understand more when I'm doing
something with the knowledge flow
application that's why I'm going
straight to the knowledge flow tab here
so if you click on knowledge flow you
can have a window just like this with a
knowledge flow environment and you can
see that we have we are having lots of
tabs here the first one starts with data
sources
we have data sings filters classifiers
cluster so these are the tabs I think
you're familiar with because these tabs
are also present in the wake and
Explorer application and we have a pin
here at the bottom that shows the status
because when you are using knowledge
flow and when you are playing with the
classifiers on datasets on clock cross
validation testing training and so on at
that time this is possible that your you
can experience some errors so those
errors if they are present in your
experimental setup they can be shown
here and you can always take a look at
those errors by clicking on the lock tab
you can see that I was having some
errors as well as I performed some
experiments previously perfectly so this
these things are locked down here so you
have the facility to take a look at what
your experiment is doing and what your
experiment is the fault your experiment
is having so to begin with you can click
on data source and here you can load any
file that Waker supports for example our
file c-45 CSV comma separated values our
database files and so on so these are
the files that are supported by Waker to
play with so as I have some of our some
of my art files that are produced during
my experiments I'll start with our
floater so when you are clicking on the
our floater at the data source tab but
when you when you click them you have a
cross here just like this so you click
on the knowledge flow layout here so
after doing that this means that this
particular option the icon here this
will load them
that you are trying to play with after
doing that we are going to evaluation
and we will choose the class assignor so
when you're clicking that will have a
crosshair again and you go to the
knowledge flow layout and click on that
so what this class assignor does is if
you can remember when I was dealing with
the Java API of Waker all the time I was
saying which of the attributes in my
attribute list is actually the class
attribute that labels the instances in
your data set so that's why I have to
choose the class assign are from the
evaluation tab and put that into the
knowledge flow layout to connect them to
connect the art floater with the class
assignor you have to right click on our
floater and you can click either
instance or data set but for the sake of
this tutorial I'll stick with the data
set only because what instance does I'll
come back to it in some later tutorials
so click on data set and you can see
that we are having some squares blue
squares around the logo of our floater
and class assignor and also we are
having a rubber band a rubber band
connector here from a generated from our
floater so we are trying to connect our
filter with the class assignor so I take
this rubber band connected to class
assignor and left-click on there so we
are having now a red arrow that says
that data set as its label and which our
file we are trying to load in order to
do that either you can double click on
this icon or you can right click on this
icon and you can select configure
so from configure I will choose the art
file that I was playing with for many
times to dot ARF
so I loaded that our file so what it
does is it loads that our file when I
will run this program run this around
the setup and it will fit that data set
to the class assignor so the now the
task of class assignor is to identify
the attribute from my attribute list
that actually labels the instances in my
data set so again I right-click on the
icon class assignor and I select
configure and from these options you can
see that these are the attributes I was
having in my to dot our file and from
those attributes I know that the nominal
attribute called
class is my class attribute so I select
that and click on OK so the next is I
have to split my dataset and I'm
actually doing the cross-validation so
in order to do that I have to stay with
the evaluation tab and you can see there
is an option called cross-validation
fold maker so what I am choosing that I
have the crosshair and clicking left
clicking on the knowledge flow layout
and the icon is there now so what I'm
doing is I took the data set I assigned
a class now I'm assigning I'm sending
that data set to the cross-validation
fold maker that will automatically
create the validation false
cross-validation false so in order to do
that I have to connect those icons
together class assignor and
cross-validation so in order to do that
I have to right click on class a signer
and I select data set because I'm
sending the data set to cross-validation
maker I have the elastic or the rubber
band connector in blue
and I take my mouse to cross-validation
fold maker icon and I left click on that
so now as I have sent my data set to
cross-validation for Baker now I have to
apply a classification algorithm in
order to classify the validation set are
instances so in order to do that we can
go to the classifier step click on that
and for this tutorial I'm going to stick
with the night based classifier I click
that and put that into the layout and to
connect them because I sent the data set
to the cross-validation fold maker the
cross-validation folds have been made
and now I have to send the training set
to the night base the same in the same
way I connected our floater with class
assignor and class assignor with
cross-validation fold maker this right
click on that and this time I'll take
the test set as well to night base
because in cross-validation you have to
consider broad training set and test set
for getting the results by inducing a
classifier so what I've done with that I
have to evaluate the classifier in order
to do that I select the evaluation tab
and there you can see you can see we
have a classifier performance evaluator
option I click on that and put that into
the layout now I'm going to join the
night base icon with the classifier
performance evaluator and to do that I
right click on not dying based
classifier and as this is the best
classifier because I'm actually having
the 10-fold cross-validation with the
whole data set so my classifiers will be
a batch classifier I select that and
again connect those to the night base
and classifier performance evaluator
so now we I am having a connector
between those two with the label best
classifier now I'm going to see or
visualize exactly the way in the same
way I visualized in the Explorer of Weka
so I click on visualization tab and as
you can remember that we can always
produced text type of results in its
output painting the classifier tab so we
have an option called text viewer I
click on that and put that into the
layout and I have to bring the results
from the classifier performance
evaluator to text viewer so I right
click on that and as those are texts so
I select on text and I select text and
connect that to the text viewer so now I
am finished I'm almost done
the to run in order to run this setup so
this is your knowledge floor layout okay
so this is the visualization of exactly
what you did with with the Explorer
application of Weka but this time you
can see because we are dealing with
dealing this in knowledge flow layout so
to run this setup I have to right click
on the our floater and there is an
action called start loading so I click
on that and you can see that the log has
been changed a lot and if you go to the
status then you can see that everything
has been performed very well or without
any error if there is an error then you
can have some highlighted red items here
if there is an own if there is any
warning then you can have those warnings
in yellow labels yellow colored labels
so at that time if you're having an
error or warning then you can go back to
the log section you can see what's
really happening there what's what's
really gone wrong so we are not having
any error for this case and we have also
loaded the our file here everything has
been done for you and the result is
waiting for you at this text viewer icon
so you can right click on this icon and
we have the option called show results
if I click on show results then you can
see that this is the exactly this result
is similar to what we have seen in the
Explorer we have the position recall F
score as well as to true positive and
false positive rates RLC carve and also
we have the confusion matrix as well as
we have some information on miss
classification and correct us correctly
classified instances of your data set so
in this case you have the options to
save this setup this is kind of picture
light right and you can always insert
this picture in your method section if
you're writing a research paper or if
you're writing a report for your
homework so in order to do that if you
want to save this as a BMP or JPEG image
you have to click on ctrl you have to
select control alter shift and
left-click then you have the option to
save this image as BMP jpg PNG or
PostScript EPS files so again for your
convenience I'm going to write down here
what I did to save this method ARBs
safety setup actually I click on call to
control control sorry
control alter and shift and then with
these keys pressed I clicked left mouse
button
and then I have the option to save the
image as BMP or JPEG okay so now I'm
saving that as one BMP and I click on
save and on my desktop I have the BMP
file you can see that this is the setup
I made for my experiment and it's been
saved as BMP file so you can always
reuse this BMP file in your report or in
the research paper you are trying to
write and there is also a functionality
to save your result as well in order to
save that result we were having with our
classifier night based classifier we go
to the text viewer and we take the
action show results so these are this is
the result we were having with my with
our classifier so again to save this we
phrase control alter shift and
left-click and we can save this as 2 dot
BMP so you can go back to your desktop
and I sort them by type so that thing
comes up so we have the 2 dot BMP and so
you can see this is the result we had
with our night based classifier on to
dot our file so that's pretty much of it
that's the that's a very good way to
start working with knowledge flow
environment to he'll be back with more
videos with other options that you can
use when you're dealing with knowledge
flow environment and I think you'll like
this environment more than a like the
Explorer environment