﻿1
00:00:00,300 --> 00:00:02,933
Hello and welcome to this art tutorial.

2
00:00:02,933 --> 00:00:05,666
Today we're going to implement
Support Vector Machine

3
00:00:05,666 --> 00:00:08,033
or more commonly called SVM.

4
00:00:08,033 --> 00:00:12,533
And after having implemented it on Python
we're going to do it on R.

5
00:00:12,900 --> 00:00:16,833
So let's quickly set the folder as Working
directory machine learning A-Z

6
00:00:17,400 --> 00:00:20,066
part three classification SVM.

7
00:00:20,066 --> 00:00:21,200
And here we are.

8
00:00:21,200 --> 00:00:24,500
Make sure that you have
the social network ads dot csv file.

9
00:00:24,900 --> 00:00:26,433
And then you can click on this more button

10
00:00:26,433 --> 00:00:29,066
here to set the folder as working
directory.

11
00:00:29,066 --> 00:00:29,966
All good.

12
00:00:29,966 --> 00:00:31,300
Now we're going to go to our

13
00:00:31,300 --> 00:00:34,833
classification template that we made
in the logistic regression section.

14
00:00:35,266 --> 00:00:40,833
So let's take everything from here
to the end.

15
00:00:41,066 --> 00:00:45,566
Copy and let's paste it in our SVM file.

16
00:00:46,233 --> 00:00:46,766
Here we go.

17
00:00:46,766 --> 00:00:49,766
And now we need to change very few things.

18
00:00:49,966 --> 00:00:52,733
So first
we need to create our classifier here

19
00:00:52,733 --> 00:00:55,666
which is going to be the SVM classifier
of course.

20
00:00:55,666 --> 00:00:59,533
And then we just need to change the title
here for the graph

21
00:00:59,533 --> 00:01:01,200
the training set results.

22
00:01:01,200 --> 00:01:04,600
So we're going to specify
that it's the SVM classifier.

23
00:01:05,066 --> 00:01:08,066
Then for the test set here SVM.

24
00:01:08,066 --> 00:01:13,066
And now let's move
back up to create our classifier okay.

25
00:01:13,066 --> 00:01:15,633
So as usual
we're going to import a library.

26
00:01:15,633 --> 00:01:19,866
And this time the library
is going to be the most popular library

27
00:01:19,866 --> 00:01:24,000
that is used for SVM
which is the library E10 71.

28
00:01:24,466 --> 00:01:25,600
So let's do this first.

29
00:01:25,600 --> 00:01:27,700
Let's have a look at the packages here.

30
00:01:27,700 --> 00:01:31,066
Check out to see if you have the E10
71 library.

31
00:01:31,500 --> 00:01:34,800
It is installed on my RCU
because I used it a lot of times, but

32
00:01:34,966 --> 00:01:37,966
if you installed R for the first time
then you might not have it.

33
00:01:38,133 --> 00:01:40,200
So I'm going to just

34
00:01:40,200 --> 00:01:43,933
write this command here
for those of you who need to install it.

35
00:01:43,933 --> 00:01:47,266
So install dot packages. Here it is.

36
00:01:47,533 --> 00:01:50,766
And then in parentheses you put in quotes

37
00:01:51,133 --> 00:01:55,500
the name of the library
which is named this way E10 71.

38
00:01:56,133 --> 00:01:56,533
All right.

39
00:01:56,533 --> 00:02:01,500
So if you select and execute this line
it will install the package.

40
00:02:01,500 --> 00:02:05,400
I won't do it because it's already
installed but it will do it I promise you.

41
00:02:05,700 --> 00:02:08,700
So I'm going to put that as comment there.

42
00:02:09,100 --> 00:02:12,333
And now let's start
creating our classifier.

43
00:02:12,333 --> 00:02:18,300
Well first we need to,
you know, add this line library E10 71.

44
00:02:18,300 --> 00:02:21,833
And that's in case
you want to make some automated script

45
00:02:21,900 --> 00:02:25,700
that select automatically
your library here because this won't be

46
00:02:25,700 --> 00:02:26,700
selected all the time.

47
00:02:26,700 --> 00:02:28,800
So you want to make sure
that you select it.

48
00:02:28,800 --> 00:02:31,900
And now we are going
to create our SVM classifier.

49
00:02:31,900 --> 00:02:34,566
So as usual
we're going to call our classifier.

50
00:02:34,566 --> 00:02:37,166
Classifier equals.

51
00:02:37,166 --> 00:02:41,566
And then here we're going to use
very simply the function SVM.

52
00:02:42,033 --> 00:02:43,866
So let's do this SVM.

53
00:02:43,866 --> 00:02:47,066
Then let's press
F1 to look at the parameters f1.

54
00:02:47,533 --> 00:02:50,366
And here are the parameters the arguments.

55
00:02:50,366 --> 00:02:52,666
So the first argument is formula.

56
00:02:52,666 --> 00:02:53,766
So let's add it.

57
00:02:53,766 --> 00:02:58,966
So as usual it's the formula
that is the dependent variable expressed

58
00:02:59,100 --> 00:03:02,600
using a tilde with respect
to all your independent variables

59
00:03:02,600 --> 00:03:04,400
which you represent by a dot.

60
00:03:04,400 --> 00:03:07,300
So here
we're just going to add formula equals.

61
00:03:07,300 --> 00:03:11,100
So first we take our dependent variable
which is purchased.

62
00:03:11,966 --> 00:03:12,900
All right.

63
00:03:12,900 --> 00:03:15,200
And now we use this tilde here.

64
00:03:15,200 --> 00:03:18,433
And then a dot a dot meaning
that we're taking

65
00:03:18,433 --> 00:03:22,400
all the independent variables of our data
set okay.

66
00:03:22,633 --> 00:03:28,000
Now comma to go to the next argument
then the next argument is data.

67
00:03:28,333 --> 00:03:32,266
So of course that's the data
on which you want to train your classifier

68
00:03:32,433 --> 00:03:34,033
on which you want your class for you

69
00:03:34,033 --> 00:03:37,233
to learn the data
to make the future classification.

70
00:03:37,833 --> 00:03:41,033
And of course, this data is our
training set.

71
00:03:41,933 --> 00:03:42,866
Okay. Perfect

72
00:03:43,833 --> 00:03:44,500
comma.

73
00:03:44,500 --> 00:03:45,700
Let's go to the next argument.

74
00:03:45,700 --> 00:03:48,633
The next argument is x and y.

75
00:03:48,633 --> 00:03:50,766
But we don't really care about that.

76
00:03:50,766 --> 00:03:55,300
And however we do
care about the type here and the kernel

77
00:03:56,166 --> 00:04:00,466
because the type as you notice
there's two type of SVM.

78
00:04:00,466 --> 00:04:05,300
There is the SVM for classification
and the SVR for regression,

79
00:04:05,300 --> 00:04:08,866
which is kind of the same support
vector machines algorithm,

80
00:04:08,866 --> 00:04:11,400
but there is one for classification
and one for regression.

81
00:04:11,400 --> 00:04:15,900
So here with this argument type
you choose the classifier version.

82
00:04:15,933 --> 00:04:19,766
And the default type for classification
is C classification.

83
00:04:19,966 --> 00:04:22,800
So that's the one
we're going to choose for type.

84
00:04:22,800 --> 00:04:25,800
So let's add here type equals

85
00:04:26,266 --> 00:04:29,266
c classification.

86
00:04:30,333 --> 00:04:31,300
All right.

87
00:04:31,300 --> 00:04:34,766
And now the last very important
argument is the kernel.

88
00:04:35,166 --> 00:04:40,800
So we're starting with the most simple SVM
which is the linear SVM.

89
00:04:40,800 --> 00:04:43,266
So here
we're going to choose the linear kernel.

90
00:04:43,266 --> 00:04:46,433
And then we're going to try
some more sophisticated SVM

91
00:04:46,433 --> 00:04:48,300
with some Gaussian kernel.

92
00:04:48,300 --> 00:04:51,066
But for now
we're going to choose a linear kernel.

93
00:04:51,066 --> 00:04:55,466
So here we'll input kernel equals linear.

94
00:04:55,733 --> 00:04:56,166
All right.

95
00:04:56,166 --> 00:05:00,066
And that's all with these four arguments
our classifier is ready.

96
00:05:00,066 --> 00:05:02,400
Our SVM classifier is ready.

97
00:05:02,400 --> 00:05:07,166
So we're going to select the different
steps or our sections of our code.

98
00:05:07,366 --> 00:05:11,133
So let's first select
this one to preprocess the data.

99
00:05:11,833 --> 00:05:14,600
Here it is okay. So perfect.

100
00:05:14,600 --> 00:05:18,000
We have our data set here
our training set and our test set.

101
00:05:18,900 --> 00:05:23,900
So the data set contains 400 observations
which are informations about users

102
00:05:23,900 --> 00:05:27,100
of a social network including the age
and the estimated salary.

103
00:05:27,100 --> 00:05:27,766
In this column.

104
00:05:27,766 --> 00:05:30,300
Purchase here tells if yes or no.

105
00:05:30,300 --> 00:05:35,000
The user bought a car
when the user got the ad that

106
00:05:35,000 --> 00:05:39,300
the car company put on the social network
for marketing campaign purposes.

107
00:05:39,600 --> 00:05:42,933
And so our class very as usual,
it's going to try to classify

108
00:05:43,300 --> 00:05:45,300
the users into the two categories.

109
00:05:45,300 --> 00:05:47,533
Yes, they bought the car
and no, they didn't buy the car.
