﻿1
00:00:00,033 --> 00:00:00,833
Hello, my friends.

2
00:00:00,833 --> 00:00:05,000
All right, let's make that prediction
with the SVR model.

3
00:00:05,000 --> 00:00:07,500
So we're going to create a new code
cell here.

4
00:00:07,500 --> 00:00:08,366
And here we go.

5
00:00:08,366 --> 00:00:11,366
We're going to start
of course from our regressor

6
00:00:11,566 --> 00:00:16,500
from which we're going to call the predict
method that will take as input.

7
00:00:16,633 --> 00:00:20,333
And now I'm asking you
what what will it exactly take as input.

8
00:00:20,533 --> 00:00:23,600
Well it's not directly the 6.5 level

9
00:00:23,700 --> 00:00:27,433
but the scaled value of the 6.5 level.

10
00:00:27,500 --> 00:00:27,866
Right.

11
00:00:27,866 --> 00:00:33,033
Because our SVR model was trained
on the scaled values at the training set,

12
00:00:33,033 --> 00:00:37,233
and therefore in the predict method,
we must enter the scaled value

13
00:00:37,233 --> 00:00:39,600
of the input that we want to predict.

14
00:00:39,600 --> 00:00:44,300
And therefore here we must call our scalar
x object

15
00:00:44,433 --> 00:00:48,600
from which
we're going to call the transform method.

16
00:00:48,766 --> 00:00:49,366
There we go.

17
00:00:49,366 --> 00:00:52,133
And then in this transform
method of our scaler object,

18
00:00:52,133 --> 00:00:55,500
well that's where we can enter
our position level of 6.5.

19
00:00:55,500 --> 00:00:59,600
But remember we have to enter it
in a double pair of square brackets

20
00:00:59,733 --> 00:01:03,466
because the predict method expects
any input as a 2D array.

21
00:01:03,733 --> 00:01:04,200
All right.

22
00:01:04,200 --> 00:01:07,800
So let's enter 6.5.
And now that's not all.

23
00:01:07,800 --> 00:01:09,833
There are two extra things we need to do.

24
00:01:09,833 --> 00:01:13,800
Remember that
we not only scaled the inputs in X

25
00:01:14,066 --> 00:01:16,666
but we also scaled the output.

26
00:01:16,666 --> 00:01:19,266
Remember
I'll go back to feature scaling here.

27
00:01:19,266 --> 00:01:22,066
We made that scaler object for the input x

28
00:01:22,066 --> 00:01:25,433
but also that scaler object
as y for the output y.

29
00:01:25,433 --> 00:01:29,966
And indeed we used that scaler
y object here to scale the output y.

30
00:01:29,966 --> 00:01:34,033
And therefore since the output
y was killed, well, in order to get

31
00:01:34,200 --> 00:01:38,066
the prediction in the original scale,
meaning the original salaries,

32
00:01:38,233 --> 00:01:43,100
well, we must apply a reverse scaling
to that whole prediction.

33
00:01:43,333 --> 00:01:43,866
Right.

34
00:01:43,866 --> 00:01:49,300
And the method that will do exactly
this is a method called inverse transform.

35
00:01:49,433 --> 00:01:54,300
And this will exactly reverse this killing
that we applied to the output y.

36
00:01:54,533 --> 00:01:56,633
All right.
So let's do this. Let's call this method.

37
00:01:56,633 --> 00:01:59,400
First
we have to call it from our c underscore

38
00:01:59,400 --> 00:02:03,433
y object because we want to reverse
the scaling of the output y.

39
00:02:03,533 --> 00:02:07,966
So let's see why from which
we call this inverse underscore.

40
00:02:08,166 --> 00:02:10,600
There it is inverse transform method.

41
00:02:10,600 --> 00:02:14,500
And then we will put this whole prediction
in the parentheses

42
00:02:14,500 --> 00:02:17,333
of this inverse transform
method of the object.

43
00:02:17,333 --> 00:02:18,300
So here we go.

44
00:02:18,300 --> 00:02:21,600
We add a parenthesis here
and we close it right here.

45
00:02:22,233 --> 00:02:23,666
All right almost ready.

46
00:02:23,666 --> 00:02:25,233
We're almost ready to get that prediction.

47
00:02:25,233 --> 00:02:27,533
But there is one less thing we need to do.

48
00:02:27,533 --> 00:02:29,600
And you don't need to worry
too much about this.

49
00:02:29,600 --> 00:02:31,800
This is just for the SVR model.

50
00:02:31,800 --> 00:02:34,600
You won't have to apply
too much of reshapes

51
00:02:34,600 --> 00:02:39,066
and all the other models of this course,
but to avoid a format error,

52
00:02:39,066 --> 00:02:42,933
we must just add inside the parentheses
of the inverse transform

53
00:02:42,933 --> 00:02:46,733
method another reshape
which is dot reshape

54
00:02:46,966 --> 00:02:50,966
and then in parentheses
you enter minus one and one.

55
00:02:51,000 --> 00:02:51,600
And this way

56
00:02:51,600 --> 00:02:55,800
we will all avoid a format error and
we'll all be able to get the prediction.

57
00:02:56,133 --> 00:02:56,700
All right.

58
00:02:56,700 --> 00:02:57,433
So that's it.

59
00:02:57,433 --> 00:02:58,800
We are ready to get the prediction.

60
00:02:58,800 --> 00:03:02,466
But first remember that in the previous
tutorial we didn't run the cell.

61
00:03:02,466 --> 00:03:05,633
So let's do it now to train the SVR model.

62
00:03:05,666 --> 00:03:06,700
Here it is.

63
00:03:06,700 --> 00:03:09,733
And so now if you're ready
let's get the predicted

64
00:03:09,733 --> 00:03:13,700
salary of the position level 6.5
by the SVR model.

65
00:03:13,866 --> 00:03:14,566
Here we go.

66
00:03:14,566 --> 00:03:21,000
Let's run this cell and we get $170,370.

67
00:03:21,333 --> 00:03:23,000
All right. It looks pretty good.

68
00:03:23,000 --> 00:03:26,633
It looks to make pretty good sense,
but that we will double check

69
00:03:26,633 --> 00:03:30,466
in the next tutorial
by visualizing the SVR results.

70
00:03:30,666 --> 00:03:31,900
Try to do it before me.

71
00:03:31,900 --> 00:03:34,766
You will again
have to play with the transform

72
00:03:34,766 --> 00:03:37,033
unless you are inverse transform.
But you can do it.

73
00:03:37,033 --> 00:03:39,300
You can of course start from the codes

74
00:03:39,300 --> 00:03:41,233
at the end of the polynomial regression
notebook.

75
00:03:41,233 --> 00:03:42,900
At least that's where we'll start from.

76
00:03:42,900 --> 00:03:44,200
So I look forward to this.

77
00:03:44,200 --> 00:03:46,200
And until then, enjoy machine learning.
