1
00:00:00,300 --> 00:00:01,200
So let's do this.

2
00:00:01,200 --> 00:00:04,100
You can not actually really guess
how we're going to do this.

3
00:00:04,100 --> 00:00:06,300
So let's implement this solution together.

4
00:00:06,300 --> 00:00:10,366
The first step is to create
an object of the column transform a class.

5
00:00:10,600 --> 00:00:15,133
And therefore we're going to create a new
variable here which I'm going to call c t

6
00:00:15,400 --> 00:00:19,500
which stands for you know the object
of the column transform a class.

7
00:00:19,800 --> 00:00:21,533
So city equals.

8
00:00:21,533 --> 00:00:25,333
And now of course remember to create
an instance or an object of a class.

9
00:00:25,566 --> 00:00:27,000
You have to call the class itself.

10
00:00:27,000 --> 00:00:29,766
So column transform or a class.

11
00:00:29,766 --> 00:00:30,900
There we go.

12
00:00:30,900 --> 00:00:33,900
Now since it is a class
we have to add some parentheses

13
00:00:33,966 --> 00:00:37,200
inside
which we have to enter two arguments.

14
00:00:37,200 --> 00:00:41,500
So these two arguments are first
transformers where we will specify

15
00:00:41,500 --> 00:00:45,633
what kind of transformation we want to do
and on which indexes the columns

16
00:00:45,633 --> 00:00:47,000
we want to transform.

17
00:00:47,000 --> 00:00:51,300
And the second argument is remainder
which will specify this.

18
00:00:51,300 --> 00:00:53,266
We actually want to keep the columns

19
00:00:53,266 --> 00:00:57,400
that won't be applied some transformations
meaning age and salary.

20
00:00:57,833 --> 00:00:58,500
So let's do this.

21
00:00:58,500 --> 00:01:01,600
Let's first enter
the names of these two arguments.

22
00:01:01,833 --> 00:01:03,266
The first one is trend

23
00:01:04,366 --> 00:01:05,400
for mirrors.

24
00:01:05,400 --> 00:01:07,933
There we go. This one Transformers.

25
00:01:07,933 --> 00:01:09,400
We're going to add a column here.

26
00:01:09,400 --> 00:01:11,033
And then let's just add the next one.

27
00:01:11,033 --> 00:01:12,300
And then we'll input them.

28
00:01:12,300 --> 00:01:14,900
The next one is read main.

29
00:01:14,900 --> 00:01:16,900
Do this one exactly.

30
00:01:16,900 --> 00:01:18,633
And now let's input them.

31
00:01:18,633 --> 00:01:23,633
So for the first one Transformers
we actually have to specify three things.

32
00:01:23,866 --> 00:01:27,066
First the kind of transformation
which is encoding.

33
00:01:27,466 --> 00:01:31,500
Second, what kind of encoding
we want to do, which is one hot encoding.

34
00:01:31,800 --> 00:01:34,733
And third, the indexes of the columns

35
00:01:34,733 --> 00:01:37,733
we want to encode meaning country
to country column.

36
00:01:38,166 --> 00:01:40,066
So we have to actually input all that

37
00:01:40,066 --> 00:01:43,766
in a pair of square brackets
and then some parenthesis.

38
00:01:44,000 --> 00:01:47,533
That's just the format expected
by this Transformers argument.

39
00:01:47,933 --> 00:01:50,133
We're actually going to enter here
a tuple.

40
00:01:50,133 --> 00:01:52,933
You know
three elements inside these parenthesis.

41
00:01:52,933 --> 00:01:55,933
The first element is, as we said,
the kind of transformation.

42
00:01:56,100 --> 00:01:59,733
And to specify that we want to do
some encoding transformation,

43
00:02:00,000 --> 00:02:03,266
we have to enter here in quotes and coder.

44
00:02:03,700 --> 00:02:05,066
So that's the first element.

45
00:02:05,066 --> 00:02:06,766
Then second element.

46
00:02:06,766 --> 00:02:08,300
We simply have to enter

47
00:02:08,300 --> 00:02:12,100
the exact name of the class
that will proceed to this encoding.

48
00:02:12,366 --> 00:02:15,600
So I'm going to copy this
because this is going

49
00:02:15,600 --> 00:02:18,966
to be exactly
the second element expected here.

50
00:02:18,966 --> 00:02:21,600
And I'm adding some parenthesis
because that's a class.

51
00:02:22,633 --> 00:02:24,966
And finally the third element inside

52
00:02:24,966 --> 00:02:29,433
this tuple of parentheses
is in a pair of new square brackets.

53
00:02:29,666 --> 00:02:34,033
The indexes of the columns
we want to apply one hot encoding.

54
00:02:34,033 --> 00:02:35,633
You know we want to transform.

55
00:02:35,633 --> 00:02:38,366
And so these indexes of course
are only one index.

56
00:02:38,366 --> 00:02:40,266
It is the index of the country column.

57
00:02:40,266 --> 00:02:43,766
The country column is the first column
of our matrix of features.

58
00:02:43,933 --> 00:02:46,933
And remember that in Python indexes
start at zero.

59
00:02:47,033 --> 00:02:49,200
Therefore this country column has an zero.

60
00:02:49,200 --> 00:02:51,300
And that's what we have to enter here.

61
00:02:51,300 --> 00:02:52,733
Only zero.

62
00:02:52,733 --> 00:02:53,200
All right.

63
00:02:53,200 --> 00:02:56,633
So that's all good for the first argument
here.

64
00:02:56,633 --> 00:03:00,900
You know Transformers the Transformers
argument is equal to all this.

65
00:03:00,900 --> 00:03:02,633
You know in square brackets.

66
00:03:02,633 --> 00:03:03,300
So great.

67
00:03:03,300 --> 00:03:06,300
And now the second argument remainder.

68
00:03:06,300 --> 00:03:09,900
So here we want to specify in quotes
the following code name

69
00:03:09,900 --> 00:03:14,633
which is base through
and which is a code name that will say

70
00:03:14,633 --> 00:03:19,133
that we indeed want to keep the columns
that won't be applied, some transformation

71
00:03:19,133 --> 00:03:22,833
that won't be one hot encoded,
which are of course age and salary.

72
00:03:23,133 --> 00:03:26,533
If we don't include this remainder equals
path through here.

73
00:03:26,700 --> 00:03:30,200
Then when we apply the transformation ex,
we will only keep the,

74
00:03:30,200 --> 00:03:33,933
you know, the first three columns
resulting from one hot encoding.

75
00:03:34,066 --> 00:03:37,600
And of course we want to keep age
and salary into our matrix of features.

76
00:03:37,600 --> 00:03:40,200
That's what it's used for okay.

77
00:03:40,200 --> 00:03:43,200
So good. Now we have our city object.

78
00:03:43,366 --> 00:03:47,533
It is of course not connected yet
to our matrix features x.

79
00:03:47,766 --> 00:03:49,800
And so that's exactly what we want to do.

80
00:03:49,800 --> 00:03:52,133
But there is actually some good news
this time.

81
00:03:52,133 --> 00:03:53,633
We don't have to do it in two steps.

82
00:03:53,633 --> 00:03:57,533
You know, by first calling a fit method
from the object to connect our object

83
00:03:57,533 --> 00:04:00,533
to the matrix of features X,
and then apply as a second set

84
00:04:00,533 --> 00:04:03,000
the transform method
to apply the transformation.

85
00:04:03,000 --> 00:04:07,266
No, this time we can do it at once
because our column transformer class

86
00:04:07,266 --> 00:04:11,366
actually has a method called
Fit transform, which will do exactly

87
00:04:11,366 --> 00:04:15,300
the process of fitting and transforming
at once at the same time, you know?

88
00:04:15,300 --> 00:04:17,966
So that's perfect. Let's use this.

89
00:04:17,966 --> 00:04:21,233
And to use this well of course
we have to call first our city object,

90
00:04:21,500 --> 00:04:24,500
from which we're going to call this fit

91
00:04:24,700 --> 00:04:28,566
transform method, which will get as input.

92
00:04:28,633 --> 00:04:32,000
Well of course X
because that's what we want to transform.

93
00:04:32,000 --> 00:04:35,700
We want to transform the matrix
of features x inside which 1 to 1.

94
00:04:35,700 --> 00:04:38,733
How to encode the country column. Perfect.

95
00:04:39,133 --> 00:04:41,966
So now two things to understand.

96
00:04:41,966 --> 00:04:46,733
Well first, this fit transform
method will of course return as output.

97
00:04:47,000 --> 00:04:48,733
The new matrix of features

98
00:04:48,733 --> 00:04:52,266
x with three columns,
one hot encoding the country column.

99
00:04:52,633 --> 00:04:55,800
And therefore,
since that's exactly what we want to get

100
00:04:55,800 --> 00:04:59,766
as the new matrix of features X,
well we're just going to update

101
00:05:00,000 --> 00:05:01,433
this new matrix of features x.

102
00:05:01,433 --> 00:05:05,900
That's why I'm adding here x equals
the result of the output of this fit

103
00:05:05,933 --> 00:05:07,133
transform method.

104
00:05:07,133 --> 00:05:09,666
But then we have to do one more thing,

105
00:05:09,666 --> 00:05:12,633
which is related
to the fact that the fit transform method

106
00:05:12,633 --> 00:05:15,933
actually doesn't return
the output as a numpy array,

107
00:05:16,166 --> 00:05:20,000
and that is absolutely converter
it to have x as a numpy array,

108
00:05:20,100 --> 00:05:23,033
because this will be expected
by the future machine

109
00:05:23,033 --> 00:05:24,933
learning models,
which we're going to build.

110
00:05:24,933 --> 00:05:27,633
You know, in order to train the future
machine learning models

111
00:05:27,633 --> 00:05:30,733
where we're going to use a train function
which is actually called fit.

112
00:05:30,933 --> 00:05:35,866
And this train function will expect
the matrix of features x as a numpy array.

113
00:05:36,066 --> 00:05:37,833
So here we want to force

114
00:05:37,833 --> 00:05:40,933
the output of this fit transform
method to be a numpy array.

115
00:05:41,133 --> 00:05:45,033
And to do this we simply need to call
well numpy first, which has a shortcut

116
00:05:45,033 --> 00:05:49,366
named np from which
we're going to call this array function,

117
00:05:49,366 --> 00:05:53,866
which will take as input exactly
the output of the fit transform method.