﻿1
00:00:00,066 --> 00:00:02,633
Hello and welcome to this art tutorial.

2
00:00:02,633 --> 00:00:06,233
So for those of you who haven't checked
the Python tutorials on kernel SVM,

3
00:00:06,233 --> 00:00:08,000
I can't wait to show you the results.

4
00:00:08,000 --> 00:00:10,900
It's going to be something
you'll definitely see how the kernel

5
00:00:10,900 --> 00:00:11,766
SVM classifier.

6
00:00:11,766 --> 00:00:13,566
You can be a powerful classifier,

7
00:00:13,566 --> 00:00:17,700
especially in our situation where the data
set is not linearly separable.

8
00:00:18,000 --> 00:00:22,533
Here you're going to see how it can manage
to overcome this non-linear separability

9
00:00:22,800 --> 00:00:26,933
and classify most of our users
of the social network correctly.

10
00:00:27,433 --> 00:00:29,966
So let's make this kernel SVM classifier

11
00:00:29,966 --> 00:00:32,966
right now and quickly look at the results.

12
00:00:32,966 --> 00:00:36,333
So as usual we're going to start
by setting our working directory.

13
00:00:36,600 --> 00:00:38,400
So right now I'm on my desktop.

14
00:00:38,400 --> 00:00:41,400
I'm going to my machine
learning A-Z folder part

15
00:00:41,400 --> 00:00:44,566
three classification section kernel SVM.

16
00:00:44,566 --> 00:00:45,933
And here we are.

17
00:00:45,933 --> 00:00:49,866
Make sure that you have this
social network add CSV file a new folder.

18
00:00:49,866 --> 00:00:52,900
And if that's the case, you ready
to click on this more button here

19
00:00:53,233 --> 00:00:56,233
to set your folder as working directory.

20
00:00:56,600 --> 00:00:58,600
Here it is all good.

21
00:00:58,600 --> 00:01:01,033
And now let's start making our model.

22
00:01:01,033 --> 00:01:04,100
So we made an awesome
classification template

23
00:01:04,100 --> 00:01:06,233
in the logistic regression section.

24
00:01:06,233 --> 00:01:08,300
So we're going to use it.

25
00:01:08,300 --> 00:01:12,100
So here we're going to take everything
from here to the end.

26
00:01:12,533 --> 00:01:17,300
Copy this
and simply paste it in the kernel SVM.

27
00:01:17,300 --> 00:01:18,500
Our file.

28
00:01:18,500 --> 00:01:23,333
And now what we only need to do
is to create our classifier here.

29
00:01:23,733 --> 00:01:25,800
Because the template is made in such a way

30
00:01:25,800 --> 00:01:29,200
to make your machine learning experience
as efficient as possible.

31
00:01:29,566 --> 00:01:32,066
So we're not going to create it now.

32
00:01:32,066 --> 00:01:36,100
First we're going to, you know, select
all the data pre-processing.

33
00:01:36,100 --> 00:01:38,300
First step to preprocess the data.

34
00:01:38,300 --> 00:01:41,300
So we're going to press Command
and Control Plus and to execute.

35
00:01:41,366 --> 00:01:42,266
And all good.

36
00:01:42,266 --> 00:01:45,100
As you can see the code executed properly.

37
00:01:45,100 --> 00:01:49,300
Now we can have a quick look at the data
set the training set and the test set.

38
00:01:49,800 --> 00:01:52,800
So we can see that
we have 400 observations

39
00:01:52,800 --> 00:01:56,600
in the data set, 300 that were selected
to go to the training set.

40
00:01:56,833 --> 00:01:59,466
And 100 selected to go to the test set.

41
00:01:59,466 --> 00:02:03,233
That's because of our 
oh point 75 split ratio here.

42
00:02:04,133 --> 00:02:07,466
And quick reminder,
this dataset is about a social network

43
00:02:07,466 --> 00:02:11,300
that contains information
about users in a social network.

44
00:02:11,366 --> 00:02:12,500
So we have their age.

45
00:02:12,500 --> 00:02:16,500
The estimated salary,
and the last column tells if yes or no.

46
00:02:16,633 --> 00:02:21,133
The users bought a product of a business
kind of this social network,

47
00:02:21,133 --> 00:02:22,566
which is a car company.

48
00:02:22,566 --> 00:02:26,033
So this car company
put ads on the social network,

49
00:02:26,300 --> 00:02:31,133
and the social network gathered those
informations to see the users reaction.

50
00:02:31,466 --> 00:02:35,266
And so zero means here
that the user didn't buy the SUV.

51
00:02:35,500 --> 00:02:38,500
And one here means
that the user bought the SUV.

52
00:02:38,666 --> 00:02:42,033
And our goal now is to make a classifier
that classifies

53
00:02:42,300 --> 00:02:44,400
those users into two categories.

54
00:02:44,400 --> 00:02:48,100
The category of users that didn't buy
the SUV, and the category of users

55
00:02:48,100 --> 00:02:49,633
that bought the SUV.

56
00:02:49,633 --> 00:02:50,466
So let's do this.

57
00:02:50,466 --> 00:02:52,800
Let's do it with kernel SVM.

58
00:02:52,800 --> 00:02:55,800
And so right now
we need to create our classifier.

59
00:02:56,633 --> 00:03:00,100
And as usual it's
going to be very intuitive and simple.

60
00:03:00,333 --> 00:03:02,266
We're going to use
the best library for this.

61
00:03:02,266 --> 00:03:06,066
And besides if you watched the SVM
tutorials you're going

62
00:03:06,066 --> 00:03:09,600
to see that we're using the same library
and the same function.

63
00:03:09,866 --> 00:03:12,466
We will just need to change
some parameters.

64
00:03:12,466 --> 00:03:13,233
So let's do this.

65
00:03:13,233 --> 00:03:13,900
For those of you

66
00:03:13,900 --> 00:03:17,500
who didn't follow the SVM tutorials,
you need to install the package.

67
00:03:17,800 --> 00:03:20,666
So to do this you type this command,

68
00:03:20,666 --> 00:03:23,666
install the packages

69
00:03:23,666 --> 00:03:27,400
and in parentheses and quotes E 1071.

70
00:03:27,700 --> 00:03:30,700
So that's the most popular package
for SVMs.

71
00:03:30,733 --> 00:03:35,200
Another very popular package is kernel
that you can play around and check it out.

72
00:03:35,466 --> 00:03:36,566
It's actually quite simple.

73
00:03:36,566 --> 00:03:40,066
It's kind of the same function, only with
slightly different parameters to input.

74
00:03:40,433 --> 00:03:42,733
But it's also a great package for SVMs.

75
00:03:42,733 --> 00:03:46,266
But here we're going to use E 1071
which is one of the most popular packages.

76
00:03:46,566 --> 00:03:51,700
And so that's for you guys who you know,
if you go to your packages, this here.

77
00:03:52,066 --> 00:03:55,333
And if you go to E 1071

78
00:03:55,900 --> 00:03:59,866
you can see that, I have this package
installed on my packages.

79
00:04:00,000 --> 00:04:01,366
It might not be the case for you.

80
00:04:01,366 --> 00:04:03,133
So that's why I'm adding this line here.

81
00:04:03,133 --> 00:04:06,266
In case you don't have the package
installed on your packages in R,

82
00:04:06,566 --> 00:04:08,433
so I won't do it right now.

83
00:04:08,433 --> 00:04:10,633
You just need to select this line
and execute.

84
00:04:10,633 --> 00:04:13,200
And it will install the package
very quickly.

85
00:04:13,200 --> 00:04:16,533
So I'm just going to put that in comment.

86
00:04:17,233 --> 00:04:21,266
I just pressed command shift
plus C and now we need to do.

87
00:04:21,266 --> 00:04:25,600
Another important thing about this library
is to you know type this common library.

88
00:04:26,333 --> 00:04:30,300
And then in parentheses
the name not in quotes of the library.

89
00:04:30,300 --> 00:04:33,300
It 1071 okay.

90
00:04:33,500 --> 00:04:34,300
And we need to add

91
00:04:34,300 --> 00:04:37,466
this line is very important
in case we have some automated scripts

92
00:04:37,733 --> 00:04:41,266
that, you know, make some kernel
SVM models today.

93
00:04:41,400 --> 00:04:42,900
So imagine you have a workflow

94
00:04:42,900 --> 00:04:46,400
and you want to include the kernel
SVM model in this workflow.

95
00:04:46,500 --> 00:04:48,800
Well you will need some automated scripts.

96
00:04:48,800 --> 00:04:52,833
So it's important to add this line
because this will automatically select

97
00:04:53,033 --> 00:04:55,433
your library. Because as you can see
here, it's not selected.

98
00:04:55,433 --> 00:04:57,966
That means it's not imported in some way.

99
00:04:57,966 --> 00:05:01,666
So you need this line that will
automatically select this package.

100
00:05:02,300 --> 00:05:03,266
Okay.

101
00:05:03,266 --> 00:05:06,000
And now we are ready
to create our classifier.

102
00:05:06,000 --> 00:05:08,866
So it's going to be the same as for SVM

103
00:05:08,866 --> 00:05:11,866
which is actually and SVM with a linear
kernel.

104
00:05:12,000 --> 00:05:13,666
Here we're not going to use a linear
kernel.

105
00:05:13,666 --> 00:05:16,966
We're going to use something else
which will be the Gaussian kernel.

106
00:05:17,400 --> 00:05:18,166
So let's do this.

107
00:05:18,166 --> 00:05:21,466
We're going to call our classifier
as usual classifier

108
00:05:22,666 --> 00:05:23,700
and then equal.

109
00:05:23,700 --> 00:05:26,966
And then
here we're going to use the SVM function

110
00:05:27,266 --> 00:05:30,266
of the E 1071 library.

111
00:05:30,366 --> 00:05:32,533
And now we need to import the parameter.

112
00:05:32,533 --> 00:05:35,000
So as usual let's have a look.

113
00:05:35,000 --> 00:05:36,133
Here it is.

114
00:05:36,133 --> 00:05:38,166
And now you need to click on this.

115
00:05:38,166 --> 00:05:41,266
And here we are
in the our documentation for SVM

116
00:05:41,466 --> 00:05:44,466
1071.
