﻿1
00:00:00,133 --> 00:00:02,633
Hello and welcome to this art tutorial.

2
00:00:02,633 --> 00:00:04,100
So today in this tutorial,

3
00:00:04,100 --> 00:00:07,566
we are ready to fit multiple linear
regression to the training set.

4
00:00:07,933 --> 00:00:11,300
So as for simple linear regression,
the first thing that we're going to do

5
00:00:11,300 --> 00:00:14,833
is to introduce
the multiple linear regressor.

6
00:00:15,200 --> 00:00:18,200
And we're going to call it regressor.

7
00:00:18,233 --> 00:00:21,733
And then as for simple linear regression
we're going to use the lm function

8
00:00:22,000 --> 00:00:23,000
here lm.

9
00:00:23,000 --> 00:00:27,333
And in parentheses
we're going to input the parameters okay.

10
00:00:27,366 --> 00:00:31,166
So let's have a look at the LM
function here by pressing F1.

11
00:00:31,266 --> 00:00:33,566
Here we go. Fitting linear models.

12
00:00:33,566 --> 00:00:34,733
Let's look at our arguments.

13
00:00:34,733 --> 00:00:37,233
So that's the same as for simple linear
regression.

14
00:00:37,233 --> 00:00:39,000
Actually the first argument is formula.

15
00:00:39,000 --> 00:00:43,466
But this time you're going to see how
we will need to change the formula syntax.

16
00:00:43,666 --> 00:00:47,200
And then the second compulsory argument
to import is the data.

17
00:00:47,433 --> 00:00:49,800
And of course
it's going to be the training set.

18
00:00:49,800 --> 00:00:52,333
So let's first input the formula.

19
00:00:52,333 --> 00:00:54,133
So formula here.

20
00:00:54,133 --> 00:00:56,466
So remember in simple linear regression

21
00:00:56,466 --> 00:01:01,100
we wrote the formula salary proportional
to years experience.

22
00:01:01,266 --> 00:01:03,533
And here it's almost going to be the same.

23
00:01:03,533 --> 00:01:07,066
But the difference is that we have several
independent variables.

24
00:01:07,300 --> 00:01:10,100
We have four independent variables.

25
00:01:10,100 --> 00:01:12,900
In simple linear regression
the dependent variable

26
00:01:12,900 --> 00:01:16,366
was proportional
to the only independent variable.

27
00:01:16,866 --> 00:01:18,633
Well here it's going to be the same.

28
00:01:18,633 --> 00:01:22,533
The profit is going to be proportional
to these independent variables.

29
00:01:23,000 --> 00:01:27,700
Only the correct way to say
it is not proportional, but is the profit

30
00:01:27,900 --> 00:01:32,166
is going to be a linear combination
of the independent variables,

31
00:01:32,700 --> 00:01:36,200
and formula equals profit

32
00:01:37,200 --> 00:01:38,733
and n.

33
00:01:38,733 --> 00:01:42,400
And here the idea is to add
all the independent variables

34
00:01:42,400 --> 00:01:44,466
separated with a plus sign.

35
00:01:44,466 --> 00:01:46,566
So let's look at them.

36
00:01:46,566 --> 00:01:49,766
It's r d spend

37
00:01:50,700 --> 00:01:53,633
I'm using the dot here
because in the data set

38
00:01:53,633 --> 00:01:56,633
we have some spaces
and I'll replace them by dots.

39
00:01:57,100 --> 00:01:59,233
So here it's dot then.

40
00:01:59,233 --> 00:02:03,600
Plus and then the other variable
which is administration

41
00:02:05,033 --> 00:02:08,000
plus marketing spend

42
00:02:08,000 --> 00:02:09,966
spend plus state.

43
00:02:09,966 --> 00:02:11,133
And that's all.

44
00:02:11,133 --> 00:02:13,900
So that's how you express the profit

45
00:02:13,900 --> 00:02:17,500
as a linear combination
of all these independent variables.

46
00:02:17,700 --> 00:02:22,200
However there is a trick to write
this formula in a much more efficient way.

47
00:02:22,633 --> 00:02:25,700
It's instead of writing
all the independent variables here,

48
00:02:26,066 --> 00:02:28,700
we can simply write a dot.

49
00:02:28,700 --> 00:02:30,200
That's what we can simply write.

50
00:02:30,200 --> 00:02:32,366
And that's the exact same equation.

51
00:02:32,366 --> 00:02:35,266
R understands
that you want to express the profit

52
00:02:35,266 --> 00:02:38,733
as a linear combination
of all the independent variables.

53
00:02:39,033 --> 00:02:42,800
So the dot here just means
all the independent variables okay.

54
00:02:42,800 --> 00:02:44,900
So that's for the first argument.
Formula equals

55
00:02:44,900 --> 00:02:48,000
profit as a linear combination
of all the independent variables.

56
00:02:48,300 --> 00:02:49,933
And the second argument is data.

57
00:02:49,933 --> 00:02:53,833
And as we said it is
of course the training set

58
00:02:54,766 --> 00:02:58,533
because we want to train
our multiple linear regression model

59
00:02:58,533 --> 00:03:02,200
on the training set and then test
the performance on the test set later.

60
00:03:02,700 --> 00:03:03,133
All right.

61
00:03:03,133 --> 00:03:07,233
So that's it for building
our multiple linear regression regressor.

62
00:03:07,433 --> 00:03:11,300
So we're going to select this and execute.

63
00:03:11,700 --> 00:03:12,333
Here we go.

64
00:03:12,333 --> 00:03:15,600
And now now
there is something very important to do.

65
00:03:15,833 --> 00:03:21,633
And this is something very practical
very useful in R it's that we are going to

66
00:03:21,666 --> 00:03:25,800
look at the informations of our regressor
that we just built here.

67
00:03:26,200 --> 00:03:28,500
And to do this it's really, really simple.

68
00:03:28,500 --> 00:03:32,100
We just need to it's actually
like simple linear regression only here.

69
00:03:32,100 --> 00:03:33,466
This time
it's going to be more interesting

70
00:03:33,466 --> 00:03:35,400
because we will have several
independent variables,

71
00:03:35,400 --> 00:03:38,400
and some of them will have a stronger
effect on the dependent variable.

72
00:03:38,600 --> 00:03:40,733
So we'll see. It's
going to be more interesting.

73
00:03:40,733 --> 00:03:43,900
And so like in simple linear regression
we're going to write a summary

74
00:03:44,566 --> 00:03:45,900
of our regressor.

75
00:03:47,100 --> 00:03:47,700
All right.

76
00:03:47,700 --> 00:03:50,266
And just press enter.

77
00:03:50,266 --> 00:03:53,266
And that's
it I'm just going to move that up.

78
00:03:54,566 --> 00:03:58,933
And here we have
all the info of our regressor okay.

79
00:03:58,933 --> 00:04:02,800
So here the first thing is just a reminder
of what the formula is.

80
00:04:02,800 --> 00:04:05,800
So the profit is expressed
as a linear combination

81
00:04:05,933 --> 00:04:07,900
of all the independent variables.

82
00:04:07,900 --> 00:04:10,900
And our regressor is trained
on the training set.

83
00:04:11,133 --> 00:04:13,266
Okay. Perfect residuals.

84
00:04:13,266 --> 00:04:16,266
We will be talking about that
at the end of this part.

85
00:04:16,433 --> 00:04:18,866
So let's not focus on that right now.

86
00:04:18,866 --> 00:04:21,066
However that's the important part.

87
00:04:21,066 --> 00:04:22,766
The coefficients here.

88
00:04:22,766 --> 00:04:24,700
That's what we need to focus on right now.

89
00:04:24,700 --> 00:04:27,900
Because as you can see it gives some info

90
00:04:27,900 --> 00:04:30,900
for each of your independent variable.

91
00:04:30,933 --> 00:04:32,533
Or by the way state two and state three.

92
00:04:32,533 --> 00:04:34,733
Here are the dummy variables.

93
00:04:34,733 --> 00:04:38,300
And that's because, you know,
I told you that the library

94
00:04:38,300 --> 00:04:40,200
takes care of everything for you.

95
00:04:40,200 --> 00:04:42,966
Not only it created
the dummy variables are

96
00:04:42,966 --> 00:04:47,266
knew that it had to create dummy variables
for the state variable, and we helped R

97
00:04:47,266 --> 00:04:51,166
to understand that because we encoded
the state variable as factors.

98
00:04:51,533 --> 00:04:56,466
So or did all the job for you
because not only he created the dummy

99
00:04:56,466 --> 00:05:00,600
variables for the state variable,
but also R didn't fall into the dummy

100
00:05:00,600 --> 00:05:03,633
variable try because as you can see, it
automatically removed

101
00:05:03,633 --> 00:05:08,700
one of the dummy variable
to avoid some redundant dependency.

102
00:05:08,866 --> 00:05:09,866
So that's perfect.

103
00:05:09,866 --> 00:05:13,833
That's the beauty of, our libraries
as well as Python libraries and Python.

104
00:05:13,833 --> 00:05:17,300
I know we specified a new section
to remove the dummy variable,

105
00:05:17,300 --> 00:05:19,266
but that was just to remind you
about the trap.

106
00:05:19,266 --> 00:05:20,566
We don't need to do it.

107
00:05:20,566 --> 00:05:22,533
R and Python take care of that for you.
