﻿1
00:00:00,100 --> 00:00:02,066
And now we're eventually
at the final round

2
00:00:02,066 --> 00:00:06,400
of our logistic regression model,
which is the fun part, because we're going

3
00:00:06,400 --> 00:00:09,733
to visualize the training set results
as well as the test set results.

4
00:00:10,200 --> 00:00:13,800
And basically, we're going to make a plot
that will represent

5
00:00:13,800 --> 00:00:16,800
everything that is happening
with our logistic regression model.

6
00:00:16,933 --> 00:00:18,566
So let's make this chart.

7
00:00:18,566 --> 00:00:20,100
I'm going to do like in Python.

8
00:00:20,100 --> 00:00:23,966
I'm going to take the code
that I've prepared and paste it here.

9
00:00:24,533 --> 00:00:27,300
I will select and execute the code.

10
00:00:27,300 --> 00:00:28,966
And then for those of you
who are interested

11
00:00:28,966 --> 00:00:32,666
in the idea behind the code,
that is how the code plots the graph.

12
00:00:32,933 --> 00:00:35,933
I will explain how the code works
at the end of this tutorial.

13
00:00:36,433 --> 00:00:38,733
Okay, so right now
I'm going to paste the code.

14
00:00:38,733 --> 00:00:39,200
Here we go.

15
00:00:39,200 --> 00:00:42,200
Then this code contains 15 lines of code.

16
00:00:42,266 --> 00:00:43,800
That was also the case in Python.

17
00:00:43,800 --> 00:00:47,933
And that's because it's the exact
same code based on the same idea.

18
00:00:48,266 --> 00:00:49,600
And I will explain this idea.

19
00:00:49,600 --> 00:00:53,033
That means I will explain how this code
works at the end of this tutorial.

20
00:00:53,700 --> 00:00:57,600
But for now let's select
this code and let's see what happens.

21
00:00:57,900 --> 00:01:00,900
So command and Control
plus enter to execute.

22
00:01:01,266 --> 00:01:02,700
And let's wait.

23
00:01:02,700 --> 00:01:04,933
It's taking a little time. That's normal.

24
00:01:04,933 --> 00:01:06,100
But it's going to plot something.

25
00:01:08,566 --> 00:01:09,766
There it is.

26
00:01:09,766 --> 00:01:11,966
There it is. That's the plot.

27
00:01:11,966 --> 00:01:12,400
Okay.

28
00:01:12,400 --> 00:01:13,566
So for those of you

29
00:01:13,566 --> 00:01:17,300
who follow the Python tutorial,
we obtain exactly the same plot.

30
00:01:17,866 --> 00:01:19,033
So let's describe this plot.

31
00:01:19,033 --> 00:01:22,033
I'm going to enlarge this.

32
00:01:22,566 --> 00:01:22,900
Okay.

33
00:01:22,900 --> 00:01:25,933
So let's analyze this graph step by step.

34
00:01:26,466 --> 00:01:29,000
First let's focus on all the points. Here.

35
00:01:29,000 --> 00:01:30,900
We can see that we have some red points.

36
00:01:30,900 --> 00:01:32,900
And some green points.

37
00:01:32,900 --> 00:01:35,600
So all these points that we see on this
graph

38
00:01:35,600 --> 00:01:38,666
are the observation
points of our training set.

39
00:01:39,066 --> 00:01:42,466
That is these are all the users
of the social network

40
00:01:42,833 --> 00:01:45,833
that were selected
to go to the training set.

41
00:01:45,833 --> 00:01:50,466
And each of these users
here is characterized by its age.

42
00:01:50,466 --> 00:01:55,033
Here on the x axis and its estimated
salary here on the y axis.

43
00:01:55,566 --> 00:01:59,366
Now we can see that
there are some red points here

44
00:01:59,766 --> 00:02:02,366
and some green points here.

45
00:02:02,366 --> 00:02:06,633
The red points are the training set
observations for which the dependent

46
00:02:06,633 --> 00:02:12,033
variable purchased is equal to zero,
and the green points are the training set.

47
00:02:12,033 --> 00:02:16,633
Observations for which the dependent
variable purchased is equal to one.

48
00:02:17,366 --> 00:02:20,866
That means that the red points here
are the users

49
00:02:20,866 --> 00:02:24,033
who didn't buy the SUV,
and the green points.

50
00:02:24,033 --> 00:02:27,300
Here are the users who bought
who actually bought the SUV.

51
00:02:27,600 --> 00:02:32,033
So now, as a first step of analysis,
let's give an interpretation

52
00:02:32,266 --> 00:02:35,066
of what we observe here with these users.

53
00:02:35,066 --> 00:02:35,366
Okay.

54
00:02:35,366 --> 00:02:41,633
So first we can see that the users
who are young with a low estimated salary.

55
00:02:41,633 --> 00:02:44,766
So these users here
actually didn't buy the SUV.

56
00:02:45,433 --> 00:02:50,766
Then if we look at the users who are older
and with a higher estimated salary,

57
00:02:50,933 --> 00:02:54,266
well, we can see that most of these users
actually bought the SUV.

58
00:02:54,400 --> 00:02:58,366
And it actually makes sense
because the SUV is more like a family car

59
00:02:58,466 --> 00:02:59,700
and therefore more interesting

60
00:02:59,700 --> 00:03:02,700
for these older users here
with a high estimated salary.

61
00:03:03,033 --> 00:03:06,033
Besides,
we can also see that some older people,

62
00:03:06,233 --> 00:03:09,833
even with a low estimated salary,
actually bought the SUV,

63
00:03:10,266 --> 00:03:12,866
because we can see that
we have some green points here

64
00:03:12,866 --> 00:03:16,166
that correspond to an age
above the average.

65
00:03:16,166 --> 00:03:17,900
The average is here,

66
00:03:17,900 --> 00:03:21,166
but an estimated salary below the average
because the average is here.

67
00:03:22,000 --> 00:03:22,900
Okay.

68
00:03:22,900 --> 00:03:26,433
So these guys, these older guys,
although they have a low estimated salary,

69
00:03:26,433 --> 00:03:29,866
actually bought the SUV, probably
because they've been saving up some money.

70
00:03:29,866 --> 00:03:32,233
Or maybe they finished
paying of their mortgage.

71
00:03:32,233 --> 00:03:36,133
I don't know, but what's for sure
is that they couldn't resist buying this

72
00:03:36,566 --> 00:03:40,666
very cool luxury
SUV offered at a ridiculously low price.

73
00:03:42,000 --> 00:03:42,566
And on the other

74
00:03:42,566 --> 00:03:46,166
hand, we can also see that
there are some young people here

75
00:03:46,333 --> 00:03:50,600
with a high estimated salary
who actually bought the SUV.

76
00:03:50,933 --> 00:03:53,033
You know, maybe
because it's a very cool SUV

77
00:03:53,033 --> 00:03:56,000
and they want to impress their friends
and take them into road trips.

78
00:03:56,000 --> 00:03:58,900
Or maybe they already have a family,
I don't know.

79
00:03:58,900 --> 00:04:01,400
Anyway, they bought the SUV.

80
00:04:01,400 --> 00:04:05,800
Actually, there are a lot of buyers,
so this must be a very cool and cheap SUV.

81
00:04:06,400 --> 00:04:09,733
Okay, and now
what is the goal of classification?

82
00:04:10,033 --> 00:04:12,233
Now we're talking machine learning.

83
00:04:12,233 --> 00:04:14,700
Why are we making some classifiers?

84
00:04:14,700 --> 00:04:16,733
And what will classifiers do.

85
00:04:16,733 --> 00:04:20,533
And least what are we trying to make them
do for this particular business problem.

86
00:04:21,100 --> 00:04:23,700
Well the goal here is to classify

87
00:04:23,700 --> 00:04:26,700
the right users into the right categories.

88
00:04:26,700 --> 00:04:29,700
That is,
we are trying to make a classifier

89
00:04:29,866 --> 00:04:32,966
that will catch the right users
into the right category,

90
00:04:33,300 --> 00:04:37,166
which are yes, they buy the SUV and no,
they don't buy the SUV.

91
00:04:37,466 --> 00:04:41,100
And we represented the way
our classifier catches these users

92
00:04:41,400 --> 00:04:44,400
by plotting
what I called the prediction regions.

93
00:04:44,633 --> 00:04:48,866
And so the prediction regions are
the two regions that we see on this graph.

94
00:04:49,200 --> 00:04:52,466
This red one here and this green one here.

95
00:04:52,733 --> 00:04:56,600
And the red prediction region
is the region where our classifier catches

96
00:04:56,600 --> 00:04:59,600
all the users that don't buy the SUV.

97
00:04:59,800 --> 00:05:01,266
And the green prediction region

98
00:05:01,266 --> 00:05:05,100
is the region where a classifier catches
all the users that buy the SUV.

99
00:05:05,400 --> 00:05:06,566
But be careful.

100
00:05:06,566 --> 00:05:11,200
This is according to the classifier
that is for each user of this

101
00:05:11,200 --> 00:05:13,266
red prediction region here.

102
00:05:13,266 --> 00:05:17,400
Our logistic regression
classifier predicts that the user doesn't

103
00:05:17,400 --> 00:05:18,833
buy the SUV.

104
00:05:18,833 --> 00:05:21,833
And for each user of this green prediction
region here,

105
00:05:21,966 --> 00:05:25,533
our classifier will predict
that the user buys the SUV.

106
00:05:25,933 --> 00:05:29,400
So that makes an awesome tool
because for each

107
00:05:29,400 --> 00:05:33,033
new user of the social network,
well, a logistic regression

108
00:05:33,033 --> 00:05:37,333
classifier will tell based on its age
and its estimated salary.

109
00:05:37,733 --> 00:05:41,100
If this user belongs to this
red prediction region here,

110
00:05:41,333 --> 00:05:44,633
and therefore doesn't buy the SUV,
or if this user belongs

111
00:05:44,633 --> 00:05:48,066
to this green prediction
region here and therefore buys the SUV.
