1
00:00:00,300 --> 00:00:04,500
So for those of you who are interested in
knowing how the code works.

2
00:00:04,666 --> 00:00:06,633
Stay with me. I will explain that.

3
00:00:06,633 --> 00:00:09,533
So I will reduce this.

4
00:00:09,533 --> 00:00:11,300
Okay.

5
00:00:11,300 --> 00:00:14,300
And let's explain this.

6
00:00:14,700 --> 00:00:15,633
Okay.

7
00:00:15,633 --> 00:00:19,700
So the idea is that we consider
each of these pixel observation points.

8
00:00:20,100 --> 00:00:24,000
As a user in the social network,
like an imaginative user.

9
00:00:24,700 --> 00:00:28,100
And so, you know, for example,
this pixel point here is not a user

10
00:00:28,100 --> 00:00:29,266
in the data set.

11
00:00:29,266 --> 00:00:35,133
But we imagine this pixel point here as
a user who has this salary and this age.

12
00:00:36,300 --> 00:00:36,666
And we

13
00:00:36,666 --> 00:00:39,666
apply our classifier on this pixel
observation points

14
00:00:39,666 --> 00:00:43,100
on this user here,
so that the classifier predicts

15
00:00:43,100 --> 00:00:46,100
if the user is going to buy yes or no,
the SUV.

16
00:00:46,800 --> 00:00:49,800
Then once
the classifier makes its prediction.

17
00:00:50,133 --> 00:00:54,300
It colonizes the pixel observation points
according to the prediction.

18
00:00:54,700 --> 00:00:57,933
So if the prediction is no this pixel user

19
00:00:57,933 --> 00:01:01,033
won by the SUV,
then it will colorize in red.

20
00:01:01,300 --> 00:01:05,766
And if the prediction is yes,
this pixel user will buy the SUV,

21
00:01:06,166 --> 00:01:09,000
then it will colorize the points in green.

22
00:01:09,000 --> 00:01:12,333
And so we apply that idea
on all the pixels in this frame,

23
00:01:12,666 --> 00:01:16,133
so that eventually our classifier
will colorize

24
00:01:16,133 --> 00:01:19,400
all the points
that he predicted as zero in red,

25
00:01:19,600 --> 00:01:22,966
and all the points
that he predicted as one in green.

26
00:01:23,766 --> 00:01:26,200
So now that you get the idea,
let's look at our code

27
00:01:26,200 --> 00:01:29,333
here and go through the steps
of completing this idea.

28
00:01:30,466 --> 00:01:31,066
Okay.

29
00:01:31,066 --> 00:01:34,466
So first I declare a set equals
training set.

30
00:01:35,000 --> 00:01:36,166
That's because, you know,

31
00:01:36,166 --> 00:01:39,166
I want to plot this graph
for the training set and the test set.

32
00:01:39,366 --> 00:01:40,666
And since we're using training

33
00:01:40,666 --> 00:01:44,233
set many times in the code,
I just replaced it by set here.

34
00:01:44,566 --> 00:01:48,500
So that's when I copy paste
the same code for the test set.

35
00:01:48,833 --> 00:01:49,966
I just need to replace test

36
00:01:49,966 --> 00:01:52,966
set by training set only here
and not in the whole code.

37
00:01:53,533 --> 00:01:55,966
So basically that's just a shortcut.

38
00:01:55,966 --> 00:01:59,400
And so first we build a grid with X
one and x two.

39
00:02:00,000 --> 00:02:03,400
So we take the minimum of the values
of the training set minus one.

40
00:02:03,400 --> 00:02:06,300
Because we don't want the points
to be squeezed in the graph.

41
00:02:06,300 --> 00:02:08,833
And same for the maximum plus one.

42
00:02:08,833 --> 00:02:13,266
So by doing this we're taking the range
of our training set observation points

43
00:02:13,533 --> 00:02:17,166
minus one and plus one so that so that
our points are not squished in the graph.

44
00:02:17,866 --> 00:02:20,700
And so we're doing this for the h column

45
00:02:20,700 --> 00:02:23,966
of the training set
and the salary column of the training set.

46
00:02:25,233 --> 00:02:27,933
Then by writing this line we are building
the grid.

47
00:02:27,933 --> 00:02:30,700
So basically with these three lines
we're making the grid here

48
00:02:30,700 --> 00:02:35,100
with all the pixel observation points
that are imaginary social network users.

49
00:02:35,933 --> 00:02:40,533
Then, since this grid set is actually a
matrix of the two columns age and salary,

50
00:02:40,700 --> 00:02:44,666
but for all the imaginary users
that are the pixel observation points,

51
00:02:44,966 --> 00:02:46,200
this is actually a matrix.

52
00:02:46,200 --> 00:02:49,733
So with this line, we just give a name
to the columns of this matrix,

53
00:02:49,733 --> 00:02:52,733
which are age and estimated salary.

54
00:02:52,833 --> 00:02:55,833
And then that's
where all the magic happens.

55
00:02:55,866 --> 00:02:59,000
Because here
that's where we use our classifier

56
00:02:59,366 --> 00:03:02,033
to predict the result

57
00:03:02,033 --> 00:03:06,900
of each of the pixel observation points
that are the imaginary pixel users.

58
00:03:07,200 --> 00:03:08,966
So we predict this.

59
00:03:08,966 --> 00:03:12,800
And then, you know, since the predict
function returns the probabilities,

60
00:03:13,000 --> 00:03:16,566
we transform it into 1 or 0 result.

61
00:03:17,200 --> 00:03:20,166
So that's why I call it y grid,
because it's the predictions

62
00:03:20,166 --> 00:03:22,200
of all the points in the grid.

63
00:03:22,200 --> 00:03:25,833
So that returns the vector of predictions
of all the points in the grid.

64
00:03:26,233 --> 00:03:29,233
And then finally we plot the whole graph.

65
00:03:29,433 --> 00:03:31,166
So in this plot we include

66
00:03:32,266 --> 00:03:35,666
all our real users and their real actions.

67
00:03:36,066 --> 00:03:39,166
Read if they didn't buy the car in green,
if they bought the car

68
00:03:39,566 --> 00:03:44,100
and we plot the predicted results
of all the pixel observation points

69
00:03:44,333 --> 00:03:47,333
that we created when we created the grid.

70
00:03:48,166 --> 00:03:49,500
So that's here.

71
00:03:49,500 --> 00:03:51,900
And so, you know,
I'm using the color spring green.

72
00:03:51,900 --> 00:03:54,900
So that is the spring green color
and tomato.

73
00:03:54,900 --> 00:03:57,366
That is the tomato color.

74
00:03:57,366 --> 00:04:00,200
And for the real points
I just use a green four which is a color

75
00:04:00,200 --> 00:04:01,866
in R and red three.

76
00:04:01,866 --> 00:04:05,566
So this color is green four
and this color is red three.

77
00:04:06,233 --> 00:04:07,066
And that's it.

78
00:04:07,066 --> 00:04:09,100
That's the idea behind this code.

79
00:04:09,100 --> 00:04:12,100
It's totally fine if you didn't understand
some things in this code.

80
00:04:12,533 --> 00:04:15,866
Because anyway, we're going to use this
code as a template for our

81
00:04:15,866 --> 00:04:17,333
next classifier.

82
00:04:17,333 --> 00:04:19,600
And so we'll just have to
copy paste this code.

83
00:04:19,600 --> 00:04:20,700
But that's all.

84
00:04:20,700 --> 00:04:23,700
So that was just for those of you
who are interested in coding,

85
00:04:23,966 --> 00:04:27,133
who are interested in
how we can use code to make such a plot.

86
00:04:27,500 --> 00:04:31,766
But the important thing to understand here
is that the logistic regression

87
00:04:31,766 --> 00:04:33,500
is a linear classifier,

88
00:04:33,500 --> 00:04:37,466
which in two dimensions
means that it's a linear separator,

89
00:04:37,466 --> 00:04:40,800
and therefore it can make some
predictions, like the predictions here

90
00:04:41,033 --> 00:04:42,133
there are incorrect.

91
00:04:42,133 --> 00:04:42,566
All right.

92
00:04:42,566 --> 00:04:44,966
So thank you for watching this
R tutorials.

93
00:04:44,966 --> 00:04:46,200
And congratulations.

94
00:04:46,200 --> 00:04:50,100
Now you know how to implement
a logistic regression model using R.

95
00:04:50,433 --> 00:04:51,133
You're going to see

96
00:04:51,133 --> 00:04:54,766
that we will need some more powerful
classifiers later in our journey.

97
00:04:55,000 --> 00:04:58,866
So I look forward to continue this journey
and until then enjoy machine learning.