1
00:00:00,100 --> 00:00:01,400
And now final step.

2
00:00:01,400 --> 00:00:05,066
You know, with this line of code,
we actually built the multiple linear

3
00:00:05,066 --> 00:00:07,900
regression model.
So we have the model already on the right.

4
00:00:07,900 --> 00:00:08,666
Now it is dumb.

5
00:00:08,666 --> 00:00:11,466
You know,
it is not trained yet on the data set.

6
00:00:11,466 --> 00:00:14,066
So now we're going to make it smart
by training.

7
00:00:14,066 --> 00:00:18,766
It's indeed on our training set
composed of remember X train and Y train.

8
00:00:19,100 --> 00:00:19,666
So there we go.

9
00:00:19,666 --> 00:00:22,666
Let's take our regressor object

10
00:00:23,600 --> 00:00:26,433
from which we're going to call the fit

11
00:00:26,433 --> 00:00:29,066
method which takes us input.

12
00:00:29,066 --> 00:00:31,033
Right. Exactly the same as before.

13
00:00:31,033 --> 00:00:34,800
First, the matrix of features x train.

14
00:00:35,266 --> 00:00:40,033
And second the dependent variable vector
y train of the training set.

15
00:00:40,033 --> 00:00:40,666
Of course.

16
00:00:40,666 --> 00:00:42,300
All right. And that's it.

17
00:00:42,300 --> 00:00:45,733
This line of code will build
our multiple linear regression model.

18
00:00:45,733 --> 00:00:49,166
And this line of code
will train it on our training set.

19
00:00:49,766 --> 00:00:50,633
All right. Perfect.

20
00:00:50,633 --> 00:00:52,666
So now let's just execute the cell.

21
00:00:52,666 --> 00:00:55,633
So let me check the cells
that we didn't execute before.

22
00:00:55,633 --> 00:00:58,400
So we executed everything up to here.

23
00:00:58,400 --> 00:01:01,666
So we still have to execute this cell

24
00:01:01,733 --> 00:01:05,100
you know to split indeed the data
set into the training set and test it.

25
00:01:05,400 --> 00:01:06,800
And now we have the training set.

26
00:01:06,800 --> 00:01:10,233
And therefore
we can proceed to this next step to train

27
00:01:10,400 --> 00:01:13,200
the multiple linear
regression model on the training set.

28
00:01:13,200 --> 00:01:16,200
So here we go. Let's do it clicking play.

29
00:01:16,200 --> 00:01:17,400
And there we go.

30
00:01:17,400 --> 00:01:22,700
We now have a fully trained
linear regression model on this data set.

31
00:01:22,700 --> 00:01:26,266
So now we have a model that was trained
to understand the correlations

32
00:01:26,266 --> 00:01:30,966
between different types of spins
by 50 strips and their profit,

33
00:01:30,966 --> 00:01:34,800
so that now investors can deploy
this model on new startups

34
00:01:35,000 --> 00:01:39,300
in order to predict what profit they will
generate based on these informations.

35
00:01:39,600 --> 00:01:42,166
All right. Perfect. So congratulations.

36
00:01:42,166 --> 00:01:44,366
Now let's recap. You know three things.

37
00:01:44,366 --> 00:01:46,600
Thanks to this amazing
linear regression class.

38
00:01:46,600 --> 00:01:49,600
You don't have to worry about the dummy
variable trap.

39
00:01:49,600 --> 00:01:53,633
You don't either have to worry
about selecting the best features

40
00:01:53,633 --> 00:01:54,300
meaning the features

41
00:01:54,300 --> 00:01:57,600
with the highest p value
or the most statistically significant.

42
00:01:57,900 --> 00:02:00,466
The linear regression
class will take care of that as well.

43
00:02:00,466 --> 00:02:04,666
And also, you know how to build and train
a multiple linear regression model

44
00:02:04,900 --> 00:02:08,733
on a data set, which you also know
how to preprocess like a pro.

45
00:02:09,000 --> 00:02:09,733
All right.

46
00:02:09,733 --> 00:02:10,866
So we're now

47
00:02:10,866 --> 00:02:14,933
going to move on to the last step
which will be to predict the test results.

48
00:02:14,933 --> 00:02:19,066
Because remember that so far
our model was trained on the training set.

49
00:02:19,233 --> 00:02:23,000
And therefore we still need to check
its performance on new observations

50
00:02:23,233 --> 00:02:25,400
which are contained in the test set.

51
00:02:25,400 --> 00:02:27,966
However, here
we have to understand something important.

52
00:02:27,966 --> 00:02:29,200
You know, compared to what we did

53
00:02:29,200 --> 00:02:32,066
in the previous section
on simple linear regression.

54
00:02:32,066 --> 00:02:35,300
This time
we actually have several features, right?

55
00:02:35,300 --> 00:02:38,300
We actually have four features
instead of one like before.

56
00:02:38,333 --> 00:02:41,400
And therefore
we can not actually plot a graph

57
00:02:41,566 --> 00:02:44,100
like we did in simple linear regression
where, you know,

58
00:02:44,100 --> 00:02:48,100
we have the features in the x axis
and the dependent variable in the y axis.

59
00:02:48,100 --> 00:02:50,000
Because simply there are four features

60
00:02:50,000 --> 00:02:51,966
we would need actually
a five dimensional graph,

61
00:02:51,966 --> 00:02:54,400
which is not possible
to visualize as humans.

62
00:02:54,400 --> 00:02:57,633
So what we're going to do instead,
you know, instead of visualizing

63
00:02:57,900 --> 00:03:00,666
the test results on a plot, well,

64
00:03:00,666 --> 00:03:03,866
we will actually display two vectors.

65
00:03:04,266 --> 00:03:08,766
The first one is the vectors
of the real profits in the test set.

66
00:03:09,200 --> 00:03:13,600
And remember that the test set actually
counts for 20% of the whole data set.

67
00:03:13,866 --> 00:03:16,766
And therefore here there are,
you know, actually

68
00:03:16,766 --> 00:03:19,766
50 observations
corresponding to the 50 strips.

69
00:03:20,000 --> 00:03:24,066
Well, 20% of 50 will be ten observations
in the test set.

70
00:03:24,366 --> 00:03:27,000
So we will actually have ten samples
in the test set.

71
00:03:27,000 --> 00:03:28,966
And we will display two vectors.

72
00:03:28,966 --> 00:03:33,433
First the vector of the ten real profits
from the test set.

73
00:03:33,433 --> 00:03:37,666
You know, the ten ground truth
and the second vector of the ten predicted

74
00:03:37,666 --> 00:03:42,000
profit of the same test set,
so that we can compare for each strip

75
00:03:42,000 --> 00:03:46,400
of the test set if our predicted profit
is close to the real profit.

76
00:03:46,400 --> 00:03:48,733
And that's how we will,
you know, evaluate our model.

77
00:03:48,733 --> 00:03:51,466
And then later on in this part,
you will actually learn

78
00:03:51,466 --> 00:03:55,366
about some evaluation techniques
to measure better

79
00:03:55,366 --> 00:03:58,900
the performance of your regression models
with some relevant metrics.

80
00:03:58,900 --> 00:04:00,933
But so far, this is how we will do it.

81
00:04:00,933 --> 00:04:03,666
We will clearly see
if you know our model performs

82
00:04:03,666 --> 00:04:05,733
well on your observations,
you know, on the test set,

83
00:04:05,733 --> 00:04:10,166
because we will clearly see if the
predictions are close to the real results.

84
00:04:10,400 --> 00:04:11,100
Okay.

85
00:04:11,100 --> 00:04:13,833
So you can try to do it yourself
at least the first step,

86
00:04:13,833 --> 00:04:16,833
which is to get
the vector of predicted profit.

87
00:04:16,833 --> 00:04:20,933
But then in order to display the two
vectors, you know, next to each other,

88
00:04:21,166 --> 00:04:24,466
well, we'll have to use some certain
tricks which are not necessarily obvious.

89
00:04:24,600 --> 00:04:26,800
So we'll do that together. All right.

90
00:04:26,800 --> 00:04:29,933
So as soon as you're ready
let's proceed to the next tutorial.

91
00:04:29,933 --> 00:04:31,766
And until then enjoy machine learning.