﻿1
00:00:00,300 --> 00:00:01,166
Hello, my friends.

2
00:00:01,166 --> 00:00:02,833
All right, let's
see if you got this right.

3
00:00:02,833 --> 00:00:05,033
Again, predicting the test. Results.

4
00:00:05,033 --> 00:00:07,833
And displaying the vector.
Of the. Predictions next.

5
00:00:07,833 --> 00:00:10,600
To the vector of the real. Results,
meaning the real.

6
00:00:10,600 --> 00:00:12,400
Purchase. Decisions.

7
00:00:12,400 --> 00:00:14,100
And I really want you to.

8
00:00:14,100 --> 00:00:16,033
Juggle with all your toolkits, you know,

9
00:00:16,033 --> 00:00:19,566
whether it is the data
preprocessing toolkit or your other.

10
00:00:19,566 --> 00:00:21,500
Machine learning models,
where you have indeed.

11
00:00:21,500 --> 00:00:23,066
Several tools inside.

12
00:00:23,066 --> 00:00:24,000
Because indeed, now.

13
00:00:24,000 --> 00:00:26,500
The tool we would like to use
is that. You know.

14
00:00:26,500 --> 00:00:27,600
A little piece of code

15
00:00:27,600 --> 00:00:31,133
that allows to display
the two vectors of predicted results.

16
00:00:31,133 --> 00:00:32,633
And real. Results.

17
00:00:32,633 --> 00:00:35,333
So what I expected you to do,
you know, to.

18
00:00:35,333 --> 00:00:37,333
Be the most efficient as you can.

19
00:00:37,333 --> 00:00:40,666
Was to go to part two regression.

20
00:00:40,933 --> 00:00:45,300
And then into the multiple linear
regression folder and then open this

21
00:00:45,300 --> 00:00:50,033
multiple linear regression implementation,
which indeed contains inside.

22
00:00:50,033 --> 00:00:52,000
So I already opened it.

23
00:00:52,000 --> 00:00:55,533
That tool
allowing to display the two vectors

24
00:00:55,533 --> 00:00:58,033
of predicted results and real results.

25
00:00:58,033 --> 00:01:00,733
I'm talking,
of course, about this tool. Right.

26
00:01:00,733 --> 00:01:02,700
We implemented it many times.

27
00:01:02,700 --> 00:01:05,333
That's why I didn't want to do it again
here in R.

28
00:01:05,333 --> 00:01:07,000
Just a. Regression implementation.

29
00:01:07,000 --> 00:01:07,833
Plus I.

30
00:01:07,833 --> 00:01:10,133
Really want to train you and incentivize.

31
00:01:10,133 --> 00:01:10,900
You to.

32
00:01:10,900 --> 00:01:12,633
Be the most efficient as you can

33
00:01:12,633 --> 00:01:16,466
by juggling between your different models
implementations.

34
00:01:16,733 --> 00:01:17,233
And so.

35
00:01:17,233 --> 00:01:20,833
If you did this, if you had the reflex
to grab that tool right here.

36
00:01:20,833 --> 00:01:23,966
In multiple linear regression
or many other models

37
00:01:23,966 --> 00:01:27,100
where we implemented that,
well congratulations.

38
00:01:27,100 --> 00:01:28,233
You did amazing.

39
00:01:28,233 --> 00:01:30,600
All right.
So here we're going to take this.

40
00:01:30,600 --> 00:01:31,133
And of.

41
00:01:31,133 --> 00:01:31,666
Course if you.

42
00:01:31,666 --> 00:01:34,600
Reimplemented it. Yourself
that's also. Amazing.

43
00:01:34,600 --> 00:01:38,700
Especially if you managed
to be more efficient than how we were.

44
00:01:38,700 --> 00:01:39,400
Able to be.

45
00:01:39,400 --> 00:01:39,900
All right.

46
00:01:39,900 --> 00:01:44,300
Because what we only have to do here
is to copy this little piece of code.

47
00:01:44,300 --> 00:01:46,633
You know that tool. And.

48
00:01:46,633 --> 00:01:47,400
Paste it.

49
00:01:47,400 --> 00:01:48,100
In a new.

50
00:01:48,100 --> 00:01:51,600
Code cell here
to predict the test results.

51
00:01:51,600 --> 00:01:53,200
Because indeed we have. All the.

52
00:01:53,200 --> 00:01:55,200
Same names here. For the vector of.

53
00:01:55,200 --> 00:01:56,000
Predictions y.

54
00:01:56,000 --> 00:01:58,800
Pred, which will be the result of the.

55
00:01:58,800 --> 00:01:59,866
Predict method.

56
00:01:59,866 --> 00:02:00,800
Applied to test.

57
00:02:00,800 --> 00:02:03,800
Set and called of course from our not.

58
00:02:03,933 --> 00:02:07,133
Regressor object but classifier object.

59
00:02:07,366 --> 00:02:08,400
There we go.

60
00:02:08,400 --> 00:02:10,700
So that's the.
First change you had to make.

61
00:02:10,700 --> 00:02:13,633
And then. Well. Since this time, you know.

62
00:02:13,633 --> 00:02:14,966
Our predicted purchase.

63
00:02:14,966 --> 00:02:17,100
Decisions
and the real purchase decisions are.

64
00:02:17,100 --> 00:02:18,766
Either zero. One.

65
00:02:18,766 --> 00:02:21,333
Well we don't need to add anything
here. To.

66
00:02:21,333 --> 00:02:24,300
You know, forced a number of decimals
after the comma to.

67
00:02:24,300 --> 00:02:25,500
Be only two.

68
00:02:25,500 --> 00:02:27,833
Right here
we're. Only dealing with integers.

69
00:02:27,833 --> 00:02:29,000
So we can remove this.

70
00:02:29,000 --> 00:02:30,300
We don't need this.

71
00:02:30,300 --> 00:02:32,166
And then final question here.

72
00:02:32,166 --> 00:02:34,200
Do we have. To change anything.

73
00:02:34,200 --> 00:02:35,833
Well absolutely. Not.

74
00:02:35,833 --> 00:02:39,700
And that's what I mean by,
you know, grabbing a tool and applying it

75
00:02:39,700 --> 00:02:43,133
on your new model
by only having to change 1 or 2 things.

76
00:02:43,133 --> 00:02:47,066
Here we only changed the name of the model
type, you know, from.

77
00:02:47,066 --> 00:02:49,200
Regressor to classifier.

78
00:02:49,200 --> 00:02:50,766
Okay. So let's check it out.

79
00:02:50,766 --> 00:02:51,900
Let's see if it works.

80
00:02:51,900 --> 00:02:53,966
Let's press play here.

81
00:02:53,966 --> 00:02:55,733
And indeed.

82
00:02:55,733 --> 00:02:58,966
We get the two vector is. Next.

83
00:02:58,966 --> 00:03:00,266
To each other with.

84
00:03:00,266 --> 00:03:01,633
First on the left.

85
00:03:01,633 --> 00:03:03,300
Your vector of. Predictions.

86
00:03:03,300 --> 00:03:05,933
You know of the.
Predicted purchase. Decisions for.

87
00:03:05,933 --> 00:03:06,400
All the.

88
00:03:06,400 --> 00:03:09,333
Customers of of. Course
the test set right.

89
00:03:09,333 --> 00:03:11,333
This was applied to excess here.

90
00:03:11,333 --> 00:03:14,066
So that's all the customers
of the test set.

91
00:03:14,066 --> 00:03:16,200
And on the right in the second column you.

92
00:03:16,200 --> 00:03:19,200
Have the real purchased decisions.

93
00:03:19,400 --> 00:03:21,900
And so here
what's interesting to see is to compare.

94
00:03:21,900 --> 00:03:22,933
The predicted purchase.

95
00:03:22,933 --> 00:03:24,566
Decisions to the real ones.

96
00:03:24,566 --> 00:03:27,233
For all. The customers in the test. Set.

97
00:03:27,233 --> 00:03:27,533
All right.

98
00:03:27,533 --> 00:03:28,933
So let's see for the first.

99
00:03:28,933 --> 00:03:30,000
Customer of the test set,

100
00:03:30,000 --> 00:03:33,766
you know remember of age 30
and estimated salary 87,000.

101
00:03:33,766 --> 00:03:34,433
Dollars.

102
00:03:34,433 --> 00:03:36,533
Well the prediction is no.

103
00:03:36,533 --> 00:03:39,466
This customer didn't buy the new SUV.

104
00:03:39,466 --> 00:03:41,900
And the real result is indeed no.

105
00:03:41,900 --> 00:03:44,666
In reality,
that. Customer didn't buy the new.

106
00:03:44,666 --> 00:03:47,666
SUV. Good. Same for the second customer.

107
00:03:47,666 --> 00:03:48,966
That customer was predicted

108
00:03:48,966 --> 00:03:52,800
not to buy that new SUV, and indeed
it did not buy the new SUV.

109
00:03:53,133 --> 00:03:56,533
Third customer this time
the third customer actually.

110
00:03:56,533 --> 00:03:58,333
Bought that new SUV.

111
00:03:58,333 --> 00:04:01,866
And our model predicted
that indeed, this new customer bought it.

112
00:04:02,133 --> 00:04:05,266
Well, it's funny, we actually have a lot
of a correct predictions.

113
00:04:05,266 --> 00:04:06,533
That's amazing. Right?

114
00:04:06,533 --> 00:04:09,600
All this so far is correct.
This is correct.

115
00:04:09,733 --> 00:04:10,333
And here.

116
00:04:10,333 --> 00:04:13,300
We go. We have our first.
Incorrect prediction.

117
00:04:13,300 --> 00:04:15,066
Here. Our logistic. Regression model.

118
00:04:15,066 --> 00:04:17,200
Predicted that this particular.

119
00:04:17,200 --> 00:04:19,066
Customer didn't buy the.

120
00:04:19,066 --> 00:04:21,700
SUV because we have a prediction
of zero here.

121
00:04:21,700 --> 00:04:23,566
But in reality that.

122
00:04:23,566 --> 00:04:26,600
Customer bought that
new amazing. SUV. Okay.

123
00:04:26,600 --> 00:04:28,800
Because the real result here is a one.

124
00:04:28,800 --> 00:04:31,200
Then here. Is correct. Correct.
And here we go.

125
00:04:31,200 --> 00:04:33,300
Another incorrect prediction

126
00:04:33,300 --> 00:04:36,900
where our model predicted again
that this customer didn't buy the.

127
00:04:36,900 --> 00:04:38,900
SUV, whereas in reality.

128
00:04:38,900 --> 00:04:41,166
That customer. But the new SUV.

129
00:04:41,166 --> 00:04:42,433
All right. And you see so.

130
00:04:42,433 --> 00:04:45,166
That looks really really good
actually we will get a.

131
00:04:45,166 --> 00:04:47,033
Very nice confusion matrix.

132
00:04:47,033 --> 00:04:48,800
I will explain very soon what it is.

133
00:04:48,800 --> 00:04:49,566
And mostly a.

134
00:04:49,566 --> 00:04:53,133
Very good accuracy because the accuracy.

135
00:04:53,333 --> 00:04:54,133
In the test set.

136
00:04:54,133 --> 00:04:54,600
Of course.

137
00:04:54,600 --> 00:04:57,600
Is simply
the number of correct predictions.

138
00:04:57,733 --> 00:04:58,966
Divided by the.

139
00:04:58,966 --> 00:05:01,666
Total number of observations
in the test set.

140
00:05:01,666 --> 00:05:04,266
And this. Is exactly
what we're about to get.

141
00:05:04,266 --> 00:05:05,033
In the test set.

142
00:05:05,033 --> 00:05:08,400
We will not only get the confusion
matrix showing.

143
00:05:08,400 --> 00:05:10,600
So there. You go.
I'm about to explain what it is.

144
00:05:10,600 --> 00:05:13,733
The confusion
matrix. Will show us exactly.

145
00:05:13,733 --> 00:05:17,300
The number of correct predictions
and the number of incorrect predictions.

146
00:05:17,400 --> 00:05:20,500
For the two cases
where the real result was zero.

147
00:05:20,500 --> 00:05:21,266
Or one to.

148
00:05:21,266 --> 00:05:22,800
Basically, we will have a nice matrix

149
00:05:22,800 --> 00:05:26,400
showing how many mistakes
and correct predictions our model made.

150
00:05:26,766 --> 00:05:29,466
And of course inside the same new.

151
00:05:29,466 --> 00:05:31,366
Step or code cell.

152
00:05:31,366 --> 00:05:32,300
We will compute.

153
00:05:32,300 --> 00:05:35,200
The accuracy and we will see what is the.

154
00:05:35,200 --> 00:05:39,266
Percentage of correct predictions
are moral made on the test. Set.

155
00:05:39,666 --> 00:05:41,766
So should I ask you to try to do it.

156
00:05:41,766 --> 00:05:44,700
On your own? Well, yes. Why not? Because,
you know, you.

157
00:05:44,700 --> 00:05:46,800
Just have to go to the API of scikit.

158
00:05:46,800 --> 00:05:48,633
Learn and. Figure out.

159
00:05:48,633 --> 00:05:51,600
How to make that confusion matrix
and how to compute the.

160
00:05:51,600 --> 00:05:53,933
Accuracy.
And now I'll just give you a little hint.

161
00:05:53,933 --> 00:05:54,300
You will.

162
00:05:54,300 --> 00:05:55,533
Have to look.

163
00:05:55,533 --> 00:05:58,966
Into the metrics module from scikit learn.
