﻿1
00:00:00,233 --> 00:00:02,400
Hello and welcome to this art tutorial.

2
00:00:02,400 --> 00:00:04,366
And now we are ready to move on
to the next step,

3
00:00:04,366 --> 00:00:07,600
which is to fit our polynomial model
to the data set.

4
00:00:07,933 --> 00:00:11,733
However, I would like to show you
how the polynomial regression

5
00:00:11,733 --> 00:00:15,200
model is a much more powerful model
for our situation,

6
00:00:15,566 --> 00:00:20,166
and the best way to show you this is to
actually compare it to a baseline model,

7
00:00:20,166 --> 00:00:23,666
like a reference based model,
which will be our linear regression model.

8
00:00:24,100 --> 00:00:26,733
That's why in this tutorial
we will build two models

9
00:00:26,733 --> 00:00:29,733
the linear regression model
and the polynomial regression model.

10
00:00:30,000 --> 00:00:31,500
And then we will compare the results.

11
00:00:31,500 --> 00:00:35,366
We will compare the graphic results
and also the predictions.

12
00:00:35,400 --> 00:00:38,400
So you'll be convinced
that the polynomial regression model

13
00:00:38,566 --> 00:00:41,433
is a much more appropriate model
for this problem.

14
00:00:41,433 --> 00:00:45,433
And the main reason for that is
that our problem is a nonlinear problem.

15
00:00:46,433 --> 00:00:46,733
Okay.

16
00:00:46,733 --> 00:00:49,400
So let's start by building this linear
model.

17
00:00:49,400 --> 00:00:52,700
It's going to be very quick
because we actually already did this.

18
00:00:52,933 --> 00:00:56,500
So you know we start
by creating our regressor that this time

19
00:00:56,500 --> 00:00:57,766
we're not going to call it regressor

20
00:00:57,766 --> 00:01:01,866
because we will build two regressors
the linear one and the polynomial one.

21
00:01:02,100 --> 00:01:04,933
So we will call it Lin rank.

22
00:01:04,933 --> 00:01:05,900
Here we go.

23
00:01:05,900 --> 00:01:09,533
And then you know later we will call
our polynomial regressor poly rank.

24
00:01:10,100 --> 00:01:11,600
So Lin rank equals.

25
00:01:11,600 --> 00:01:15,300
And then remember
we have to use the LM formula.

26
00:01:15,300 --> 00:01:17,533
And so we need to add
some parentheses here.

27
00:01:17,533 --> 00:01:21,533
And then let's remind ourselves
what we need to input

28
00:01:21,800 --> 00:01:25,400
by pressing F1 here
to have a look at the arguments okay.

29
00:01:25,400 --> 00:01:28,033
So remember the first argument is formula.

30
00:01:28,033 --> 00:01:33,800
So the formula is very simply the salary.

31
00:01:33,800 --> 00:01:35,233
That is our dependent variable

32
00:01:36,266 --> 00:01:36,800
tilde.

33
00:01:36,800 --> 00:01:40,566
So I just pressed alt n
and then a dot here

34
00:01:40,566 --> 00:01:42,200
to take all the independent variables.

35
00:01:42,200 --> 00:01:45,133
But actually there is only one
independent variable.

36
00:01:45,133 --> 00:01:50,566
So the dot here is strictly equivalent
as just writing level okay.

37
00:01:50,566 --> 00:01:52,100
So perfect for the first argument.

38
00:01:52,100 --> 00:01:55,066
And then what is the second argument.
It's data.

39
00:01:55,066 --> 00:01:56,100
So data okay.

40
00:01:56,100 --> 00:02:00,900
So let's add this argument data equals
and now let's see okay.

41
00:02:00,900 --> 00:02:05,133
So in the linear regression section
we actually used the training

42
00:02:05,133 --> 00:02:09,566
set here as the data argument
to train our linear regression model.

43
00:02:09,833 --> 00:02:10,800
But as we explained

44
00:02:10,800 --> 00:02:14,466
before here we did not split the data
set into the training set or the test set.

45
00:02:14,700 --> 00:02:19,166
So we are simply going to train our model
on the whole data set itself,

46
00:02:19,400 --> 00:02:21,966
because it's a small data set anyway,
and we want to have

47
00:02:21,966 --> 00:02:24,100
the most accurate prediction.

48
00:02:24,100 --> 00:02:27,000
So here
simply we will just input data set.

49
00:02:28,400 --> 00:02:29,100
All right.

50
00:02:29,100 --> 00:02:31,633
And our linear regression model is ready.

51
00:02:31,633 --> 00:02:33,466
We are actually ready to build it.

52
00:02:33,466 --> 00:02:34,933
So let's just do it.

53
00:02:34,933 --> 00:02:37,633
Let's build our linear regression model

54
00:02:37,633 --> 00:02:40,266
I'm pressing Command and Control
plus enter to execute.

55
00:02:40,266 --> 00:02:42,666
All right linear regression made.

56
00:02:42,666 --> 00:02:45,666
We can have a quick
look at the summary here.

57
00:02:45,833 --> 00:02:48,833
Summary linear lag.

58
00:02:49,066 --> 00:02:52,133
So I'm typing this in the console
and pressing enter.

59
00:02:52,600 --> 00:02:55,566
And here are the statistical results
of the model.

60
00:02:55,566 --> 00:02:58,933
We're doing this on R because
as you can see it's really easy on Python.

61
00:02:58,933 --> 00:03:01,900
We would have needed to import
a class create an object here.

62
00:03:01,900 --> 00:03:02,533
It's really easy.

63
00:03:02,533 --> 00:03:04,866
So we can have a look and we can see that

64
00:03:04,866 --> 00:03:08,433
the level has two stars here
for the level of significance

65
00:03:08,700 --> 00:03:12,166
and is actually not
a bad predictor of the salary.

66
00:03:12,300 --> 00:03:14,200
But wait for the polynomial model

67
00:03:14,200 --> 00:03:17,500
to see how it's going
to be much better than linear regression.

68
00:03:18,166 --> 00:03:20,500
Okay,
so now let's move on to the next level.

69
00:03:20,500 --> 00:03:23,500
That is the better model
polynomial regression.

70
00:03:23,633 --> 00:03:25,466
And let's build it okay.

71
00:03:25,466 --> 00:03:30,733
So as I mentioned we are going to call
this regressor poly underscore rank.

72
00:03:30,900 --> 00:03:32,100
All right this way.

73
00:03:32,100 --> 00:03:34,666
And you know
since the polynomial regression model

74
00:03:34,666 --> 00:03:39,633
is actually a multiple linear regression
model in which the independent variables

75
00:03:39,633 --> 00:03:42,900
are actually the polynomial features
of one independent

76
00:03:42,900 --> 00:03:45,900
variables, as Carol
explained in the intuition tutorial.

77
00:03:46,033 --> 00:03:50,566
Well, we are going to use again the alarm
function as we did for linear regression.

78
00:03:50,900 --> 00:03:54,033
So here I'm just going to start
by taking my lm function.

79
00:03:54,200 --> 00:03:57,600
Add the parenthesis
and we're going to input our two arguments

80
00:03:57,900 --> 00:04:01,100
formula equals salary

81
00:04:02,033 --> 00:04:05,466
tilde of n and then this dot here.

82
00:04:05,800 --> 00:04:08,666
But don't worry
we'll actually represent something else.

83
00:04:08,666 --> 00:04:11,566
So so far I'm just putting the dot
and you'll understand

84
00:04:11,566 --> 00:04:15,133
what's going to happen next
which will make our polynomial regression

85
00:04:15,300 --> 00:04:17,000
built correctly.

86
00:04:17,000 --> 00:04:19,766
So then comma
and then we add our second argument

87
00:04:19,766 --> 00:04:22,700
which is still going to
be the data equals data set.

88
00:04:23,700 --> 00:04:24,733
Because we're going to train

89
00:04:24,733 --> 00:04:28,033
our polynomial regression
model on the whole data set okay.

90
00:04:28,533 --> 00:04:30,433
So now you must be telling yourself wait.

91
00:04:30,433 --> 00:04:33,066
But this is exactly the same
as linear regression.

92
00:04:33,066 --> 00:04:33,800
Well that's true.

93
00:04:33,800 --> 00:04:36,833
And that's why
we need to add a little something here

94
00:04:36,833 --> 00:04:39,900
to make this model
a polynomial regression model.
