﻿1
00:00:00,300 --> 00:00:02,733
Hello and welcome to this art tutorial.

2
00:00:02,733 --> 00:00:07,066
In the previous tutorial was a lot of,
interesting things to learn about R

3
00:00:07,066 --> 00:00:11,033
and what it can do to provide insightful
informations of your model.

4
00:00:11,500 --> 00:00:15,900
And now what we only have to do
left is to predict the test results.

5
00:00:16,300 --> 00:00:19,300
As for a simple linear regression,
it will be very quick.

6
00:00:19,366 --> 00:00:22,066
We will just need one line
for this and it's going to be

7
00:00:23,066 --> 00:00:23,766
y pred.

8
00:00:23,766 --> 00:00:26,833
I'm introducing
the vector of predictions here Y pred.

9
00:00:27,833 --> 00:00:30,833
And then I'm going to use
the predict function.

10
00:00:31,633 --> 00:00:34,200
And then the predict function
I will just use two arguments.

11
00:00:34,200 --> 00:00:36,233
One is the regressor.

12
00:00:36,233 --> 00:00:38,733
You have to specify with which

13
00:00:38,733 --> 00:00:41,733
regressor
you want to predict your test set results.

14
00:00:41,966 --> 00:00:44,000
And of course it's
with our multiple linear

15
00:00:44,000 --> 00:00:47,000
regression regressor that we define here.

16
00:00:47,833 --> 00:00:50,433
And the second argument is new data

17
00:00:50,433 --> 00:00:53,433
which is the set of the new observations.

18
00:00:53,433 --> 00:00:56,433
You want to predict the result
you want to predict the profit.

19
00:00:56,900 --> 00:00:59,600
And so of course this new data

20
00:00:59,600 --> 00:01:02,600
is going to be the test set.

21
00:01:03,566 --> 00:01:04,866
Don't worry if so far

22
00:01:04,866 --> 00:01:07,866
we are only making predictions
about the test set.

23
00:01:07,933 --> 00:01:11,366
You will see that in the next sections
we will be making predictions

24
00:01:11,366 --> 00:01:12,966
of new observations,

25
00:01:12,966 --> 00:01:16,500
which won't be the test set observations,
but some single observations.

26
00:01:16,800 --> 00:01:19,533
You will understand why we will do this,
and you will see that

27
00:01:19,533 --> 00:01:22,666
the purpose of doing
that is actually going to be pretty fun.

28
00:01:23,466 --> 00:01:25,566
So the okay, so that's ready.

29
00:01:25,566 --> 00:01:26,900
The predict function is ready.

30
00:01:26,900 --> 00:01:29,900
It has its two argument
the regressor and the new data.

31
00:01:29,900 --> 00:01:34,100
So we're ready to select this
and execute.

32
00:01:34,400 --> 00:01:35,400
All right.

33
00:01:35,400 --> 00:01:39,733
And now our widespread vector
of predicted results of the test

34
00:01:39,733 --> 00:01:41,500
set is ready.

35
00:01:41,500 --> 00:01:45,300
So what I'm going to do now is to type

36
00:01:45,300 --> 00:01:48,300
Y pred in the console
to have a look at it.

37
00:01:48,600 --> 00:01:49,933
Here it is.

38
00:01:49,933 --> 00:01:53,566
So as you can see this
contains ten predicted results.

39
00:01:53,566 --> 00:01:57,800
These are the ten predicted profits
of the ten observations of the test set.

40
00:01:58,633 --> 00:02:03,266
So for example let's compare
the real profits and the predicted profit.

41
00:02:03,666 --> 00:02:06,766
So let's look at our test set here okay.

42
00:02:06,766 --> 00:02:09,233
So let's look at our startups.

43
00:02:09,233 --> 00:02:10,733
Let's look at startup number four.

44
00:02:10,733 --> 00:02:14,800
And startup
number four has a $182,000 profit.

45
00:02:15,300 --> 00:02:19,333
And what about the prediction
$173,000 profit.

46
00:02:19,766 --> 00:02:23,333
So that's the predicted profit
which is not far from the real profit.

47
00:02:23,333 --> 00:02:25,500
So that's pretty good okay.

48
00:02:25,500 --> 00:02:26,533
What about the second one.

49
00:02:26,533 --> 00:02:32,633
166,000 for the real profit and 172,000
for the predicted profit.

50
00:02:32,833 --> 00:02:34,800
Not too bad either.

51
00:02:34,800 --> 00:02:38,466
Then third observation 155 160. Good

52
00:02:39,800 --> 00:02:42,166
146 135.

53
00:02:42,166 --> 00:02:44,400
Not too bad then. Okay.

54
00:02:44,400 --> 00:02:46,766
Then. For example,
let's look at the start.

55
00:02:46,766 --> 00:02:52,233
Number 24 started
number 24 had a real profit of $108,000,

56
00:02:52,600 --> 00:02:55,700
and our predicted profit is $110,000.

57
00:02:55,733 --> 00:02:58,400
So here we're pretty close.
That's an accurate prediction.

58
00:02:58,400 --> 00:03:00,566
Good one okay.

59
00:03:00,566 --> 00:03:05,066
And oh that's even better for the last
two ones because here it's 99,000.

60
00:03:05,100 --> 00:03:07,733
Here it's almost 99,000.

61
00:03:07,733 --> 00:03:11,500
And 97 here
for the last one and almost 97.

62
00:03:11,833 --> 00:03:14,000
So that's pretty good.
That's a good model.

63
00:03:14,000 --> 00:03:19,066
And remember that's the model
with all the independent variables.

64
00:03:19,366 --> 00:03:22,566
But as we saw
there is only truly significant variable

65
00:03:22,566 --> 00:03:25,033
which was the R&D spend.

66
00:03:25,033 --> 00:03:29,066
And so if we try to recompute
the regressor by only putting the R&D

67
00:03:29,066 --> 00:03:32,400
spending dependent variable here,
you will see that the predictions

68
00:03:32,400 --> 00:03:35,033
are going to be the same
with the same accuracy,

69
00:03:35,033 --> 00:03:40,066
because the R&D spend is the only strong
predictor of the profit.

70
00:03:40,400 --> 00:03:41,666
So you can try that yourself.

71
00:03:41,666 --> 00:03:43,066
It could be a good practice.

72
00:03:43,066 --> 00:03:46,066
You can play around with the different
independent variables.

73
00:03:46,266 --> 00:03:49,500
So I will let you do it
yourself and continue to analyze,

74
00:03:49,500 --> 00:03:52,500
interpret and play around
with the multiple linear regression.

75
00:03:52,866 --> 00:03:54,300
That's the end of this tutorial.

76
00:03:54,300 --> 00:03:58,900
I was very happy to teach you how to make
multiple linear regression model.

77
00:03:59,000 --> 00:04:02,133
I look forward to seeing you
in the next section where we'll be talking

78
00:04:02,133 --> 00:04:05,400
about a new kind of regressor,
because this time you're going to see that

79
00:04:05,400 --> 00:04:09,666
it's not a linear regressor,
it's going to be a polynomial regressor.

80
00:04:09,966 --> 00:04:13,200
And I can't wait to show you
how it can give powerful predictions.

81
00:04:13,533 --> 00:04:16,066
And we are going to use it
on a very funny example.

82
00:04:16,066 --> 00:04:18,200
So there will be quite a challenge here.

83
00:04:18,200 --> 00:04:20,666
And just to give you a teasing
of what we are going to do,

84
00:04:20,666 --> 00:04:22,766
we are going to make
kind of a lying detector.

85
00:04:22,766 --> 00:04:24,866
So I look forward to seeing you
in the next section.

86
00:04:24,866 --> 00:04:26,733
And until then, enjoy machine learning.
