﻿1
00:00:00,133 --> 00:00:00,966
Hello my friends.

2
00:00:00,966 --> 00:00:03,233
Welcome to this new practical activity.

3
00:00:03,233 --> 00:00:06,733
On this time,
the Naive Bayes classification model,

4
00:00:06,900 --> 00:00:12,300
we already implemented logistic regression
k nearest neighbors, SVM and kernel SVM.

5
00:00:12,500 --> 00:00:15,866
And so far the best accuracy
we've got with these four models

6
00:00:16,066 --> 00:00:19,433
is oh point 93, which we got with both

7
00:00:19,566 --> 00:00:22,566
K-nearest neighbors and kernel SVM.

8
00:00:22,600 --> 00:00:25,900
And now we're about to implement
a new classification model,

9
00:00:25,900 --> 00:00:28,966
the Naive Bayes
one, a very well known and widely used.

10
00:00:29,100 --> 00:00:31,466
So you definitely want
to have it in a toolkit.

11
00:00:31,466 --> 00:00:34,366
And so the big question is will we beat

12
00:00:34,366 --> 00:00:37,366
that best accuracy of 0.03.

13
00:00:37,466 --> 00:00:38,600
And now let's do this.

14
00:00:38,600 --> 00:00:42,700
Let's start
the implementation of Naive Bayes.

15
00:00:42,700 --> 00:00:44,300
So we're going to go into this folder

16
00:00:44,300 --> 00:00:46,300
and we're going to start with Python
as usual.

17
00:00:46,300 --> 00:00:49,166
And inside this folder
you will find two files

18
00:00:49,166 --> 00:00:52,166
the Naive Bayes
implementation in the Ipynb format

19
00:00:52,333 --> 00:00:57,700
and the same social network as data set,
which I will quickly re-explain.

20
00:00:57,700 --> 00:01:00,866
So this is a dataset
containing 400 customers.

21
00:01:00,866 --> 00:01:04,366
Each row here represents the customer,
and for each customer

22
00:01:04,400 --> 00:01:07,733
we get the two features age
and estimated salary.

23
00:01:07,833 --> 00:01:12,000
With which we are going to predict
the dependent variable purchased,

24
00:01:12,200 --> 00:01:16,500
which tells if yes or no, the customer
but the SUV.

25
00:01:16,500 --> 00:01:19,733
So zero means that indeed the customer
didn't buy the SUV,

26
00:01:19,733 --> 00:01:22,733
and one means that the customer
but the SUV.

27
00:01:22,733 --> 00:01:25,933
All right, so we're going to train
a new classification model to understand

28
00:01:26,100 --> 00:01:26,866
the correlations

29
00:01:26,866 --> 00:01:31,100
between these two features age and salary,
and that dependent variable purchased

30
00:01:31,200 --> 00:01:35,133
in order to predict the customers
that will buy the SUV.

31
00:01:35,133 --> 00:01:38,066
And once we get these predictions,
we will target these customers

32
00:01:38,066 --> 00:01:41,733
on social networks with some beautiful
ad of this beautiful car.

33
00:01:42,066 --> 00:01:43,900
All right, so that's the same story.

34
00:01:43,900 --> 00:01:47,500
And now we're going
to start our implementation.

35
00:01:48,000 --> 00:01:51,333
And I can't wait to see
if we're going to beat the accuracy.

36
00:01:51,333 --> 00:01:55,333
And I can also to show you
the visualization results at the end.

37
00:01:55,333 --> 00:01:55,600
All right.

38
00:01:55,600 --> 00:01:58,566
So right now
it is loading and laying out the notebook.

39
00:01:58,566 --> 00:01:59,400
And here it is.

40
00:01:59,400 --> 00:02:01,366
Here is a Naive Bayes implementation

41
00:02:01,366 --> 00:02:04,500
still resulting
from the same classification template

42
00:02:04,500 --> 00:02:07,033
which we made in the first section
of this part three.

43
00:02:07,033 --> 00:02:09,066
When implementing logistic regression.

44
00:02:09,066 --> 00:02:11,933
So basically all the cells here
are the same

45
00:02:11,933 --> 00:02:16,166
as in this logistic regression
implementation or classification template.

46
00:02:16,333 --> 00:02:20,500
The only cell that will change will be,
you know, the one where we build

47
00:02:20,700 --> 00:02:24,066
and train the classification model,
meaning this one.

48
00:02:24,333 --> 00:02:26,466
That's the only cell that we will

49
00:02:26,466 --> 00:02:29,600
re-implement ourselves
because all the rest is the same.

50
00:02:29,600 --> 00:02:32,533
But this notebook is in read only mode.

51
00:02:32,533 --> 00:02:36,000
And therefore, in order to re-implement
that cell, we need to create

52
00:02:36,000 --> 00:02:40,800
a copy of this notebook by clicking
save a Copy in Drive.

53
00:02:40,800 --> 00:02:42,900
And as you can see,
this will create a copy

54
00:02:42,900 --> 00:02:46,800
in which we will be able to re-implement
that cell.

55
00:02:47,266 --> 00:02:47,666
All right.

56
00:02:47,666 --> 00:02:50,100
So once again laying out the notebook
and there we go.

57
00:02:50,100 --> 00:02:54,866
Here is our copy on which
we are authorized to modify anything.

58
00:02:54,866 --> 00:02:58,433
And especially that cell
we want to re implement.

59
00:02:58,433 --> 00:03:00,733
So I'm scrolling down here to find it.

60
00:03:00,733 --> 00:03:04,500
Here it is training the naive base model
on the training set.

61
00:03:04,500 --> 00:03:06,200
So let's immediately

62
00:03:06,200 --> 00:03:10,233
remove that cell because I want you
to re-implement it from scratch.

63
00:03:10,233 --> 00:03:13,166
As if we had no idea
on how to implement this.

64
00:03:13,166 --> 00:03:15,166
All right, let's create a new code cell.

65
00:03:15,166 --> 00:03:17,066
And now your turn.

66
00:03:17,066 --> 00:03:22,600
I want you to train your machine learning
independence by figuring out by yourself

67
00:03:22,600 --> 00:03:27,200
how to indeed build and train
that naive base model on the training set.

68
00:03:27,866 --> 00:03:30,000
So of course, to do this,
you have several options.

69
00:03:30,000 --> 00:03:33,600
You can directly type
in the search bar of Google or Bing.

70
00:03:33,866 --> 00:03:37,033
Well, Naive Bayes scikit learn class okay.

71
00:03:37,033 --> 00:03:41,466
Or you can navigate the cycling API
which is right here.

72
00:03:41,700 --> 00:03:46,066
In order to find that class which
we need to build our Naive Bayes model.

73
00:03:46,400 --> 00:03:50,100
And my personal recommendation
is to try it with the second option,

74
00:03:50,100 --> 00:03:53,766
because indeed, this
will get you familiar with this API.

75
00:03:53,766 --> 00:03:56,033
And the more you get familiar
with it, the better.

76
00:03:56,033 --> 00:03:57,600
Okay, so please press pause.

77
00:03:57,600 --> 00:03:59,400
Now try to find this class

78
00:03:59,400 --> 00:04:02,866
and then try to build this naive base
model in training on the training set.

79
00:04:03,166 --> 00:04:07,000
And now in two seconds
we will implement the solution together.

80
00:04:08,000 --> 00:04:09,500
All right let's do this.

81
00:04:09,500 --> 00:04:12,500
Let's build
this naive base model with scikit learn.

82
00:04:12,866 --> 00:04:14,566
So the API is huge.

83
00:04:14,566 --> 00:04:17,366
And this is organized
in alphabetical order.

84
00:04:17,366 --> 00:04:21,300
And so the first thing we'll try
is to find, you know, a module called

85
00:04:21,300 --> 00:04:23,200
maybe Naive Bayes. Right.

86
00:04:23,200 --> 00:04:24,600
Just the name of the model.

87
00:04:24,600 --> 00:04:28,533
So we will scroll down down to N,
you know the letter N

88
00:04:28,800 --> 00:04:33,600
and we'll see if we get Naive Bayes
somewhere or something close to it.

89
00:04:33,600 --> 00:04:34,200
Let's see.

90
00:04:34,200 --> 00:04:38,566
Linear model meaningful learning metrics
we're getting close.

91
00:04:38,566 --> 00:04:42,900
And Gaussian mixture models
model selection multiclass.

92
00:04:42,900 --> 00:04:47,100
Well there is a lot of n okay
perfect naive Bayes.

93
00:04:47,100 --> 00:04:49,200
So let's click this Naive Bayes.

94
00:04:49,200 --> 00:04:49,800
All right.

95
00:04:49,800 --> 00:04:53,233
And so the model indeed
is naive underscore Bayes.

96
00:04:53,500 --> 00:04:54,733
And among these models.

97
00:04:54,733 --> 00:04:57,600
Well according to which one
are we going to take here.

98
00:04:57,600 --> 00:04:59,033
That's a good question actually.

99
00:04:59,033 --> 00:05:03,100
Well in fact the classic one and the one
that you learned in the intuition

100
00:05:03,100 --> 00:05:07,700
lecture is of course, this one,
the Gaussian, naive Bayes, Gaussian and B.

101
00:05:07,733 --> 00:05:13,000
And that's exactly what we'll use
to implement our Naive Bayes model.

102
00:05:13,566 --> 00:05:13,833
All right.

103
00:05:13,833 --> 00:05:16,066
So the name of the class is Gaussian B.

104
00:05:16,066 --> 00:05:17,033
And this time

105
00:05:17,033 --> 00:05:18,900
the good news is
that we won't have to worry

106
00:05:18,900 --> 00:05:23,566
too much about the parameters because
there are actually only two parameters.

107
00:05:23,866 --> 00:05:26,000
And so here very simply

108
00:05:26,000 --> 00:05:30,200
we will just call this class
without inputting any parameters.

109
00:05:30,366 --> 00:05:31,800
So that's super easy.

110
00:05:31,800 --> 00:05:37,266
Let's copy this and let's
go back to our implementation to copy one.

111
00:05:37,733 --> 00:05:39,533
Let's paste that here.

112
00:05:39,533 --> 00:05:41,700
Let's remove this little thing here.

113
00:05:41,700 --> 00:05:43,800
And now you know how to adapt this.

114
00:05:43,800 --> 00:05:45,833
We need to start with from,

115
00:05:45,833 --> 00:05:49,666
you know, the Naive Bayes
module of the scikit learn library.

116
00:05:49,933 --> 00:05:52,933
And then here you add import.

117
00:05:53,400 --> 00:05:56,533
There we go that Gaussian and B class.
