Weka Tutorial 23: Classification 101 using API (Classification) Video Lecture | Weka Tutorial - Data & Analytics

hello youtube and welcome to another
wicked tutorial so in this tutorial I'm
going to show you how with just a single
click you can train a classifier from
their training data and you can apply
that classifier on unseen test instances
to classify them I'm showing you this
because I've already shown you how to do
these kind of things with Explorer but
in Explorer you know it's a bit
difficult for you to use Explorer to do
these stuffs because you have to upload
the training file then you have to
create a model you have to save it and
so on
so I'm just doing it using the Java API
in this case and I'm using the Java API
to reduce the whole thing whole machine
learning idea the basic machine learning
idea of training and testing unseen
instances with a single click so Java
API is a very powerful tool developed by
the wake up team so I'm just going to
use that so before going into this
tutorial let me show you my training and
test data set so this one is my training
data set is called iris training but I
named that to to dot ARF
you can see that in this data set I have
total four attributes one two three four
and I have a class attribute called
class which has which can actually have
three types of very values there now if
we are going back to my test set then
you can see that in lighter set I have
again four attributes there and these
four attribute name and data type they
has they have to be exactly the same
name and data type as in your training
data set and also I have the class
attribute in the very end which contains
the exact number and exact number of
values as in your training data set and
their spelling has to be the same the
difference here between the training and
test data set for this example we are
just assuming that our test instances
are not labeled yet so that's why we are
developing we are inducing a classifier
from training data and we will apply
that classifier on these instances all
those instances and then my classifier
will tell what value is in there so this
is the basic idea and the just two
things you have to remember that your
training and test instances attribute
number numbers numbers there they have
to be the same their name has to be the
same the data data type has to be the
same and also your class attribute has
to contain the exact same number of
values as in your training data and
their spelling just be careful with the
spelling as well otherwise you'll get an
error okay now I'm going back to the
Java code I have written for this
tutorial so as I'm going to classify I'm
just naming that class of my of update
of these app class as classified so I
will go through every line of this code
first I have created a bufferedreader
object to read my training and test
dataset at first I'm reading it with
reading reading light winning data with
it so my training data is here to dot
ARF
then I'm creating a training object of
type instances and I'm giving the
training data set to that training train
object and I'm saying that okay as you
have my training data set remember
whether that my class index has is is my
last attribute as you have seen that
also the class attribute of my to dot or
file is sitting at the very
and I'm using the same bufferedreader
here you can use a different one but I
think it's a good idea to use the same
one and I'm reading the 1 dot R which is
my test test data set and again I'm
creating an object called test of type
instances and I'm providing my test
dataset to that particular object and
again I'm saying that my class index is
my last attribute in my attributes list
and then I close the bufferedreader
there okay
now I'm sorry having some technical
difficulties here ok now as we go along
in this tutorial I'm just going to
induce classifier from j48 algorithm of
course you can have you have the
facility to use the others like knife is
SVM or random forests so these options I
just commented it out because they vary
from classifier algorithm to algorithm
so I just don't think it's a good idea
because it's a general kind of tutorial
so I just left it out
and this j48 type of object tree I'm
using this tree to build a classifier
from the training data so it's very very
straightforward thing and before doing
anything else I'm creating another
object called labeled of type instances
so this will contain all the instances
attesting senses after they are labeled
by my classifier j48 classifier and this
labelled object will have my test set
because they are going to be the same
the only difference is my labeled
objects will contain the class
attributes provided by my classifier
then I have a loop for loop because I
need
you iterate over all the test instances
test instance number zero to the number
of instances so every time I'm looping
through this I'll call that gen 48
objects classify instance method to
classify the first instance so now you
have four attributes just tell me the
last one and if you have the last one
put it into a double type of a variable
called cels label or class label and
then I am saying that okay as you got
that set it as my class value and then
I'm using a buffered writer when when my
for loop is done for all the test
instances I classify them with my j48
classifier I am using a profit writer
object to write it down as a label dot
ARF in my drive and then I'm just
writing it down and I also close that
buffered writer and it's not seen here
so that's pretty much a fit if we go
back to our code and run it so it's run
so we are going back to my H Drive and
here a label dot our file has been
created so if I open that now you can
see that in my wondered our file I had
all those question marks because I
didn't know before applying j48 what the
class label of these instances are all
of these instances are now as I ran that
Java file I have the closing instances
so this is the exact same test file of
one dot or file except that I have now
the test instances classes so that's
pretty much every right I believe that
is going to be very very helpful helpful
for you because by Java API using Java
API you've seen that with a single click
you are loading the datasets you are
training your classifier with the
training data
you are applying that model on the test
data set which is unseen and the labels
of the test instances are unknown you
get all of those class labels and you're
writing that straight away into a file
so this is very helpful I hope you
enjoyed this video and you are getting
used to the power of Wicca to do these
things especially the power of Java API
of Weka
so if you have any questions comments
don't hesitate to put it in the comment
section if you want to email me then you
can also communicate with me in that way
thank you very much