﻿1
00:00:00,066 --> 00:00:00,800
Hello my friends.

2
00:00:00,800 --> 00:00:03,100
Congratulations again
for having implemented

3
00:00:03,100 --> 00:00:06,100
your first classification
model, logistic regression.

4
00:00:06,300 --> 00:00:09,133
And now,
before we move on to the next model,

5
00:00:09,133 --> 00:00:12,966
I would like to make a classification
template that will allow us

6
00:00:12,966 --> 00:00:17,800
to build much more efficiently the next
classification models of this part.

7
00:00:18,100 --> 00:00:20,066
And not only that, it is a template

8
00:00:20,066 --> 00:00:23,933
that you will also be able to use
for your personal data set.

9
00:00:24,100 --> 00:00:24,733
All right.

10
00:00:24,733 --> 00:00:25,666
So let's do this.

11
00:00:25,666 --> 00:00:29,600
At the top of our studio we're going to
click file and then click New File.

12
00:00:29,600 --> 00:00:31,800
And then our script. All right.

13
00:00:31,800 --> 00:00:36,566
And we're going to save this
as classification template.

14
00:00:36,866 --> 00:00:41,266
All right then we click save and you will
see it appear in your file here.

15
00:00:41,300 --> 00:00:43,300
That's the new file we just created.

16
00:00:43,300 --> 00:00:46,800
All right then we're going to take
our logistic regression implementation.

17
00:00:46,800 --> 00:00:47,200
Again.

18
00:00:47,200 --> 00:00:48,733
We're going to select everything

19
00:00:48,733 --> 00:00:53,166
copy it and then paste it
in our classification template.

20
00:00:53,166 --> 00:00:55,033
And then let's change the title.

21
00:00:55,033 --> 00:00:58,766
We're going to replace this
by classification template.

22
00:00:59,300 --> 00:01:02,100
And then in the data
preprocessing sections here we're going

23
00:01:02,100 --> 00:01:05,400
to keep everything as is except one thing.

24
00:01:05,700 --> 00:01:09,966
We're going to add one line of code
that will make sure the classes of

25
00:01:09,966 --> 00:01:14,300
our dependent variable are indeed classes
and, you know, not continuous numbers.

26
00:01:14,500 --> 00:01:17,700
So for most of the models
like logistic regression k

27
00:01:17,733 --> 00:01:22,800
and then naive Bayes, SVM and kernel SVM,
the package that we'll use to build

28
00:01:22,800 --> 00:01:26,433
those models will recognize the classes
of our data sets as classes.

29
00:01:26,633 --> 00:01:30,033
But there will be two models
that will build, which are decision tree

30
00:01:30,033 --> 00:01:33,566
and random forest
that will not recognize them as classes.

31
00:01:33,833 --> 00:01:37,600
And therefore, to make this template
applicable to all the models

32
00:01:37,600 --> 00:01:41,366
we're going to build in this part three,
we just need to add this little code

33
00:01:41,366 --> 00:01:48,133
section here that I'm going to call, 
and coding the target feature as factor.

34
00:01:48,300 --> 00:01:51,600
And that will make sure that the purchased

35
00:01:51,900 --> 00:01:54,800
dependent variable of our data

36
00:01:54,800 --> 00:01:57,800
set will be included as factors.

37
00:01:57,800 --> 00:02:02,300
And to do so we use the factor
function inside which we enter

38
00:02:02,300 --> 00:02:06,266
first the dependent variable column
which we access from our data set.

39
00:02:06,400 --> 00:02:07,800
Then adding a dollar again.

40
00:02:07,800 --> 00:02:10,033
And then there we go purchased.

41
00:02:10,033 --> 00:02:12,400
So that's the first argument
of this factor function.

42
00:02:12,400 --> 00:02:16,500
And then we just need to specify
the two classes here with this argument

43
00:02:16,500 --> 00:02:20,466
levels that will be equal
to our two classes zero and one,

44
00:02:20,466 --> 00:02:25,800
which we have to enter in this C function
which will take as input zero and one.

45
00:02:26,133 --> 00:02:28,300
All right. So let me recap this.

46
00:02:28,300 --> 00:02:32,466
This line of code is optional
for the logistic regression model.

47
00:02:32,466 --> 00:02:35,033
The k nearest neighbor model
which will implement next.

48
00:02:35,033 --> 00:02:38,833
Then the SVM model the kernel SVM model
and the Naive Bayes model.

49
00:02:39,100 --> 00:02:42,733
And it is compulsory for the decision tree
classification model

50
00:02:42,733 --> 00:02:45,166
and for the random forest
classification model.

51
00:02:45,166 --> 00:02:45,733
All right.

52
00:02:45,733 --> 00:02:48,900
So I just want to include it here
so that we can apply this template

53
00:02:49,133 --> 00:02:51,600
to all the classification models
of this part tree.

54
00:02:51,600 --> 00:02:53,866
And for that
let's make the other required changes.

55
00:02:53,866 --> 00:02:55,433
So here when splitting the data

56
00:02:55,433 --> 00:02:58,566
set into training set and a test set,
we don't have to change anything.

57
00:02:58,800 --> 00:03:00,900
Same we can keep feature scaling.

58
00:03:00,900 --> 00:03:02,766
And then that's
where we will change some things.

59
00:03:02,766 --> 00:03:04,933
So first we will remove all this

60
00:03:04,933 --> 00:03:07,300
because that's
where we'll build our next classifier.

61
00:03:07,300 --> 00:03:12,800
And so we can add a and saying
create your classifier here okay.

62
00:03:13,000 --> 00:03:16,766
Then we can replace logistic regression
here by classifier.

63
00:03:17,333 --> 00:03:17,666
All right.

64
00:03:17,666 --> 00:03:20,533
To make it generic
then here very important.

65
00:03:20,533 --> 00:03:23,866
The logistic regression
model is the only model of this part three

66
00:03:23,866 --> 00:03:27,500
that returns first its predictions
in the form of probabilities.

67
00:03:27,666 --> 00:03:30,866
All the other models will return
directly 0 or 1.

68
00:03:31,133 --> 00:03:33,600
And therefore we're going to remove this
line.

69
00:03:33,600 --> 00:03:37,633
We're going to replace here
prop pred by y bread.

70
00:03:37,933 --> 00:03:41,000
And here in the predict function
we're going to keep classifier.

71
00:03:41,000 --> 00:03:44,400
Of course we're going to remove type
equals response.

72
00:03:44,700 --> 00:03:47,766
And we're going to keep new data equals
to set minus three.

73
00:03:48,000 --> 00:03:48,400
All right.

74
00:03:48,400 --> 00:03:49,466
And this line of code

75
00:03:49,466 --> 00:03:53,233
will be applicable to all the remaining
classification models of part three.

76
00:03:53,333 --> 00:03:56,133
We won't have to change anything here
okay.

77
00:03:56,133 --> 00:03:58,033
And then same for the confusion matrix.

78
00:03:58,033 --> 00:04:02,433
This is only for the logistic regression
model because it returns probabilities.

79
00:04:02,633 --> 00:04:05,933
And so to adapt this to a model
that returns only zero one,

80
00:04:05,933 --> 00:04:09,100
well of course
we need to remove larger than 0.5.

81
00:04:09,133 --> 00:04:09,566
Right.

82
00:04:09,566 --> 00:04:13,933
Because the widespread of our future
classification models will only contain

83
00:04:13,933 --> 00:04:17,966
zero one exactly as the dependent
variable column of the test set.

84
00:04:18,566 --> 00:04:19,300
Okay, great.

85
00:04:19,300 --> 00:04:23,300
And then finally in the last two sections
here visualizing the training set

86
00:04:23,300 --> 00:04:24,500
and test results.

87
00:04:24,500 --> 00:04:24,900
Same.

88
00:04:24,900 --> 00:04:27,666
We're going to remove the probabilities
here.

89
00:04:27,666 --> 00:04:29,533
Then cut this way grid.

90
00:04:29,533 --> 00:04:33,300
And then replace this proposed variable
by y grid.

91
00:04:33,633 --> 00:04:37,733
And then remove the type
equals response parameter so that.

92
00:04:37,733 --> 00:04:39,200
Now we don't have our predictions

93
00:04:39,200 --> 00:04:42,200
returned in the form of probabilities
but zeros and ones.

94
00:04:42,500 --> 00:04:45,966
And then of course we're going to make it
generic again by replacing logistic

95
00:04:45,966 --> 00:04:49,033
regression here by classifier okay.

96
00:04:49,033 --> 00:04:49,600
Excellent.

97
00:04:49,600 --> 00:04:53,433
So now let's copy this because
we're going to do this same down there

98
00:04:53,566 --> 00:04:56,533
for the visualization of the test result.

99
00:04:56,533 --> 00:05:00,900
So I'm going to replace these two lines
by what we just copied.

100
00:05:00,900 --> 00:05:01,600
There we go.

101
00:05:01,600 --> 00:05:05,500
And then replace logistic regression again
by classifier.

102
00:05:05,700 --> 00:05:09,633
And now congratulations
we have a generic template

103
00:05:09,633 --> 00:05:12,833
that we will be able to use
to build very efficiently.

104
00:05:12,833 --> 00:05:15,600
Our next classification
rules of board three.

105
00:05:15,600 --> 00:05:16,933
So let's save this.

106
00:05:16,933 --> 00:05:19,900
And now look forward
to building those next models with you.

107
00:05:19,900 --> 00:05:22,133
And until then enjoy machine learning.
