﻿1
00:00:00,066 --> 00:00:00,900
Perfect.

2
00:00:00,900 --> 00:00:03,766
And now you know what
the next natural step is.

3
00:00:03,766 --> 00:00:07,566
Well, it is, of course,
to create a new object of this

4
00:00:07,566 --> 00:00:09,333
random Forest Regressor class.

5
00:00:09,333 --> 00:00:13,100
And we're going to call this object
as usual regressor,

6
00:00:13,433 --> 00:00:18,500
which will be equal to, well,
you know, an instance of this class.

7
00:00:18,533 --> 00:00:23,300
That's why I'm copying it and basing it
here and adding some parentheses.

8
00:00:23,833 --> 00:00:24,366
All right.

9
00:00:24,366 --> 00:00:28,900
So this time do you think
we have to input something in this class.

10
00:00:28,900 --> 00:00:33,266
Well let's not fall into the trap
of too much easiness.

11
00:00:33,266 --> 00:00:36,400
Let's say remember that
for decision tree regression

12
00:00:36,600 --> 00:00:39,933
we actually did not input,
you know, an essential parameter.

13
00:00:39,933 --> 00:00:43,466
We just input a random state factor
in order to fix the seeds

14
00:00:43,500 --> 00:00:46,733
so that we can all have the same results
displayed in the output.

15
00:00:47,033 --> 00:00:51,833
But this time there's actually a parameter
that is very important.

16
00:00:51,833 --> 00:00:54,800
And that parameter is the number of trees.

17
00:00:54,800 --> 00:00:57,233
So that's exactly what we're going
to input here.

18
00:00:57,233 --> 00:01:01,700
The name of that parameter
is an underscore estimators.

19
00:01:02,466 --> 00:01:06,800
We're going to set it
equal to 1010 trees right ten estimators.

20
00:01:06,800 --> 00:01:09,533
Each tree is an estimate okay.

21
00:01:09,533 --> 00:01:14,166
And then as usual we're going to add
the random underscore state parameter.

22
00:01:14,166 --> 00:01:18,033
And once again we'll set that equal
to zero so that we can fix the seeds

23
00:01:18,033 --> 00:01:21,600
and get the same output
displayed in our notebook.

24
00:01:22,200 --> 00:01:23,100
Okay great.

25
00:01:23,100 --> 00:01:24,333
And now final step.

26
00:01:24,333 --> 00:01:26,166
You know this step by heart.

27
00:01:26,166 --> 00:01:28,300
Now it is so obvious for you.

28
00:01:28,300 --> 00:01:34,133
It is of course to fit the regressor
object to the whole data set.

29
00:01:34,133 --> 00:01:34,833
In other words,

30
00:01:34,833 --> 00:01:38,666
that means we are training the regressor
on the whole data set, right?

31
00:01:38,666 --> 00:01:42,500
So here we have to add a dot
and then fit applied

32
00:01:42,500 --> 00:01:45,500
to x and y.

33
00:01:45,600 --> 00:01:46,466
Perfect.

34
00:01:46,466 --> 00:01:50,033
So and that's all we had to do
you know for this random

35
00:01:50,033 --> 00:01:51,400
forest regression implementation.

36
00:01:51,400 --> 00:01:54,866
All the rest is
the same as the decision tree.

37
00:01:54,866 --> 00:01:59,700
So we can just execute all the cells here
and observe comfortably in our chair

38
00:01:59,700 --> 00:02:01,533
the final result. So let's have a look.

39
00:02:01,533 --> 00:02:04,000
I just want to warn you again
that it's not going to be pretty.

40
00:02:04,000 --> 00:02:07,266
And that's of course for the exact same
reason as decision trees.

41
00:02:07,266 --> 00:02:11,266
A random forest regression
model is way better adapted to high

42
00:02:11,266 --> 00:02:14,400
dimensional data sets or, you know, data
sets with multiple features,

43
00:02:14,400 --> 00:02:18,600
which you will see in the final section
of this part two regression,

44
00:02:18,800 --> 00:02:22,500
when not only learning
how to evaluate your regression models,

45
00:02:22,500 --> 00:02:26,800
but also on
how to select the best model for any data

46
00:02:26,800 --> 00:02:30,600
set you know for a particular data
set, you're working with all right.

47
00:02:30,600 --> 00:02:31,500
So let's do this.

48
00:02:31,500 --> 00:02:34,500
Let's execute each of these cells

49
00:02:34,500 --> 00:02:37,500
starting with importing the libraries.

50
00:02:37,500 --> 00:02:42,433
Then I'm not going to forget
I almost forgot to upload the data set.

51
00:02:42,566 --> 00:02:46,333
You know I was about to execute the cell,
but that would have returned an error

52
00:02:46,466 --> 00:02:49,033
because indeed the data
set needs to be uploaded.

53
00:02:49,033 --> 00:02:51,266
So let's upload it now. All right.

54
00:02:51,266 --> 00:02:53,666
So as usual
we have to go into our machine learning.

55
00:02:53,666 --> 00:02:56,600
It is that folder.
Wherever you put it on your machine.

56
00:02:56,600 --> 00:02:58,400
Then forward to regression.

57
00:02:58,400 --> 00:03:01,266
Then section
nine and last section of this part

58
00:03:01,266 --> 00:03:04,266
two random forest regression and Python.

59
00:03:04,300 --> 00:03:05,966
And there we go. Position salaries.

60
00:03:05,966 --> 00:03:09,666
This is still of course
the same data set okay.

61
00:03:09,666 --> 00:03:11,066
So that's all good.

62
00:03:11,066 --> 00:03:12,266
Now we have the data sets.

63
00:03:12,266 --> 00:03:15,133
And now we can import it
inside the notebook.

64
00:03:15,133 --> 00:03:18,133
Well actually inside our Python program.

65
00:03:18,300 --> 00:03:20,000
And now we're going to

66
00:03:20,000 --> 00:03:23,700
train the random forest regression model
on the whole data set.

67
00:03:24,000 --> 00:03:25,400
So let's do this.

68
00:03:25,400 --> 00:03:30,266
And this will output the random Forest
regressor model with all the parameters.

69
00:03:30,266 --> 00:03:34,533
And at this stage the only parameter
that I recommend to tune

70
00:03:34,533 --> 00:03:39,033
is indeed that number of estimators
which we chose to be equal to ten.

71
00:03:39,333 --> 00:03:41,133
All right.
And don't worry too much about the rest.

72
00:03:41,133 --> 00:03:44,300
Now this will already gives you
an excellent model.

73
00:03:44,733 --> 00:03:47,866
And then now let's predict
the final result

74
00:03:48,233 --> 00:03:52,466
by, you know, just calling this
predict message from our regressor object.

75
00:03:52,466 --> 00:03:54,966
And the predict
method just has to take as input.

76
00:03:54,966 --> 00:03:59,366
Well that's position level number 6.5
which remember

77
00:03:59,366 --> 00:04:02,533
you have to input in a double pair
of square brackets.

78
00:04:02,700 --> 00:04:07,000
Because the predict method
expects a 2D array as its input.

79
00:04:07,300 --> 00:04:09,166
Right.
So that's very important for you to know.

80
00:04:09,166 --> 00:04:10,700
But we saw this many times.

81
00:04:10,700 --> 00:04:14,000
I'm sure it has become
also very obvious for you.

82
00:04:14,233 --> 00:04:17,066
So let's do this.
Let's get this prediction.

83
00:04:17,066 --> 00:04:22,966
And we get wow we get a pretty good
prediction actually $167,000,

84
00:04:22,966 --> 00:04:26,266
which is very close to,
you know, that salary that this person

85
00:04:26,533 --> 00:04:30,933
mentioned to earn in the previous company,
which was 160 K.

86
00:04:30,933 --> 00:04:32,633
So that's very, very good.

87
00:04:32,633 --> 00:04:35,133
And now let's visualize the final result.

88
00:04:35,133 --> 00:04:38,333
Oh I actually forgot
to delete that output.

89
00:04:38,333 --> 00:04:39,200
But that's fine.

90
00:04:39,200 --> 00:04:42,200
Let's run this cell
and we'll get that output again.

91
00:04:42,500 --> 00:04:47,500
And here is the regression curve
of the random forest regression model.

92
00:04:47,500 --> 00:04:51,366
And of course it looks like a lot
the one of the decision trees.

93
00:04:51,366 --> 00:04:54,900
Although this time
there are more steps of the stairs,

94
00:04:54,900 --> 00:04:57,600
you know, remember
if the decision tree regression model,

95
00:04:57,600 --> 00:05:00,766
we had a step for each of the position
level.

96
00:05:00,933 --> 00:05:05,000
And here for example we have two steps
between two position levels.

97
00:05:05,000 --> 00:05:08,066
So that's of course
because we have more trees this time

98
00:05:08,066 --> 00:05:11,500
and therefore more splits of,
you know, the features where you have

99
00:05:11,700 --> 00:05:15,166
the same prediction, you know,
the average of the predicted salary.

100
00:05:15,433 --> 00:05:18,433
So it only makes sense
that there are more steps.

101
00:05:18,733 --> 00:05:20,933
All right. So congratulations.

102
00:05:20,933 --> 00:05:24,433
That was your final regression
model of part two.

103
00:05:24,733 --> 00:05:26,300
You now have a complete

104
00:05:26,300 --> 00:05:30,166
toolkit of regression models,
which gives you a lot of different options

105
00:05:30,166 --> 00:05:34,000
and solutions for your future data
sets and future machine learning problems.

106
00:05:34,266 --> 00:05:37,266
So I'm really happy for you
that we built together this toolkit.

107
00:05:37,366 --> 00:05:39,000
Make sure to use it the right way.

108
00:05:39,000 --> 00:05:41,700
And now we will finish
with this last section

109
00:05:41,700 --> 00:05:45,000
to teach you how to use this regression
toolkit the best way

110
00:05:45,000 --> 00:05:48,833
you know by selecting the best model
for any data set.

111
00:05:49,233 --> 00:05:51,666
All right,
so join me in this next section.

112
00:05:51,666 --> 00:05:53,433
I can't wait to see you again.

113
00:05:53,433 --> 00:05:55,300
And until then, enjoy machine learning.
