1
00:00:00,520 --> 00:00:07,650
All before going deep dive into this session, let's have a quick recap of what we have done in all

2
00:00:07,690 --> 00:00:08,330
our sessions.

3
00:00:08,940 --> 00:00:13,920
So doing lots of analysis, doing lots of techniques of feature engineering as well.

4
00:00:13,930 --> 00:00:16,850
And we have found some meaningful insight as well.

5
00:00:17,130 --> 00:00:23,640
And for machine learning aspect, what we have done, we have performed this correlation, doing feature

6
00:00:23,640 --> 00:00:28,270
encoding techniques on our data and how to handle our as well.

7
00:00:28,290 --> 00:00:34,700
What we have done in the previous session by just defining dysfunction in which we are seeing just apply

8
00:00:34,710 --> 00:00:35,850
log on the data.

9
00:00:35,860 --> 00:00:39,260
That's what we have done in our previous sessions in the session.

10
00:00:39,270 --> 00:00:45,720
What we have to do this is exactly this assignment for this particular session in which I have to apply

11
00:00:46,020 --> 00:00:55,080
techniques of feature importance on our data to select what are exactly my most important features to

12
00:00:55,080 --> 00:00:57,110
build a machine learning model.

13
00:00:57,240 --> 00:01:01,300
Because if you will observe here, you have plenty of features.

14
00:01:01,440 --> 00:01:02,400
Look at that here.

15
00:01:02,400 --> 00:01:03,720
You have tons of features.

16
00:01:04,230 --> 00:01:11,430
So from this end number of features, you have to select some subset of features that are going to contribute

17
00:01:11,430 --> 00:01:13,430
more to your machine learning model.

18
00:01:13,830 --> 00:01:17,530
That's what exactly my feature important technique will do.

19
00:01:18,000 --> 00:01:21,840
So first, let me check if I have any value in my data frame or not.

20
00:01:21,960 --> 00:01:24,320
I'm just going to call this is null.

21
00:01:24,570 --> 00:01:30,870
And let me just call to some to get a summation of all the missing values and you will figure out here

22
00:01:30,870 --> 00:01:33,730
I have the single missing value so you can just drop it.

23
00:01:33,750 --> 00:01:34,320
That's it.

24
00:01:34,560 --> 00:01:41,430
So I'm going to say data frame don't drop any and just place in place equal to truth so that you can

25
00:01:41,430 --> 00:01:44,160
obliterate every that just executed.

26
00:01:44,160 --> 00:01:51,480
And if now what you have to do now, you need your independent feature as well as your dependent features.

27
00:01:51,720 --> 00:01:56,420
So let me just define two variables as lie.

28
00:01:56,730 --> 00:01:58,860
So why is exactly your dependent feature?

29
00:01:58,870 --> 00:02:04,200
So I'm going to say data frame of is underscore cancel.

30
00:02:04,440 --> 00:02:11,220
So this is exactly your dependent feature that you have to predict with respect to different number

31
00:02:11,220 --> 00:02:12,360
of independent features.

32
00:02:12,780 --> 00:02:16,800
So now how you have to consider your independent features.

33
00:02:16,830 --> 00:02:19,340
So if let's say let me show you a thing.

34
00:02:19,350 --> 00:02:28,070
So if I'm going to say that a frame got drop and here if I'm going to set this is underscore cancel

35
00:02:28,320 --> 00:02:34,670
and if I'm not going to set any in place parameter, it means I'm not going to update the frame.

36
00:02:34,860 --> 00:02:37,060
And if I'm going to that, you want to sell.

37
00:02:37,080 --> 00:02:42,270
But before executing, you have to set one more parameter, which is exactly exis.

38
00:02:42,390 --> 00:02:44,700
It means how you have to drop this column.

39
00:02:44,970 --> 00:02:52,470
So if I'm going to execute it now, you will see here you have all your features except that this one,

40
00:02:52,860 --> 00:03:00,500
it means this is your data frame or you can say this is that set that exactly your independent feature.

41
00:03:00,510 --> 00:03:03,680
So I'm going to say let me just copy from here.

42
00:03:03,960 --> 00:03:04,870
Let me just did it.

43
00:03:05,130 --> 00:03:07,930
And in this X I have to just assigned it.

44
00:03:07,950 --> 00:03:08,590
That's it.

45
00:03:08,880 --> 00:03:10,830
Now just aggregate it.

46
00:03:10,830 --> 00:03:17,190
Now you have to perform your future importance concept on your data for this.

47
00:03:17,190 --> 00:03:23,540
You have to import some libraries that will definitely help you to do all these several task.

48
00:03:24,060 --> 00:03:31,800
So for this, I'm going to say from this cyclone module, I have something which is exactly my linear

49
00:03:31,800 --> 00:03:32,720
and it's called model.

50
00:03:32,730 --> 00:03:38,970
And from this I have to import my algorithm, which is exactly my Lessel, which is exactly this one.

51
00:03:39,540 --> 00:03:46,510
After what we have to do, let's say from this ascalon, I have something which is exactly my feature

52
00:03:46,510 --> 00:03:51,210
and it's called selection to select your most important features.

53
00:03:51,240 --> 00:03:55,620
This is exactly that some module that is going to help you a lot.

54
00:03:55,650 --> 00:04:01,110
So from this module, you have to import something known as select and just step.

55
00:04:01,110 --> 00:04:09,100
And here this is a class that is going to help you a lot to perform this feature importance concept.

56
00:04:09,120 --> 00:04:14,040
So here what I have to do very first, I have to initialize this select from model.

57
00:04:14,040 --> 00:04:19,430
And here what I have to do, I have to specify my this last regression model.

58
00:04:19,650 --> 00:04:23,040
So here what I'm going to do, I'm just going to say initialize this.

59
00:04:23,220 --> 00:04:26,330
And here I have two very first, Manchanda Slessor.

60
00:04:26,700 --> 00:04:31,350
And if I will press schist crosstab over here, you will get all Erastus.

61
00:04:31,590 --> 00:04:37,380
What exactly is the alphabet amedure and what is your fitness going to set all these custom parameters

62
00:04:37,580 --> 00:04:37,770
here?

63
00:04:38,250 --> 00:04:42,240
So the main parameter that you have to take care of it with is exactly this.

64
00:04:42,240 --> 00:04:45,390
Elsah So what exactly this Alphie is.

65
00:04:45,390 --> 00:04:48,330
So it is equivalent to your penalty parameter.

66
00:04:48,390 --> 00:04:54,210
It means the bigger the value of the alpha, the less number of features will get selected.

67
00:04:54,360 --> 00:04:54,690
Let's see.

68
00:04:54,690 --> 00:04:56,550
I just need more number of features.

69
00:04:56,560 --> 00:04:59,550
It means I have to set some low value of.

70
00:04:59,910 --> 00:05:05,520
So here I'm going to say Alpha, let's say close to zero point zero zero five, let's say I'm going

71
00:05:05,520 --> 00:05:12,480
to set this value from my own experience, or if you want to customize this value, you can just something

72
00:05:12,480 --> 00:05:14,310
known as cross-validation.

73
00:05:14,490 --> 00:05:19,620
But here I am going to set this threshold value from my own experience, working on different, different

74
00:05:19,620 --> 00:05:21,690
project on different different domain.

75
00:05:21,990 --> 00:05:28,530
So after it, what I have said, I have to set something known as my random parameter and assign its

76
00:05:28,620 --> 00:05:32,710
value as zero after doing all this stuff.

77
00:05:33,000 --> 00:05:41,070
What I have to do then I have to use this select from model, which is exactly inside my cyclotron.

78
00:05:41,070 --> 00:05:47,530
And this tells how this will select the features whose coefficients are non-zero.

79
00:05:47,550 --> 00:05:50,920
That's what that's what the select phone model will do.

80
00:05:51,180 --> 00:05:54,740
So here I am going to say it's object, like I'm going to store it somewhere else.

81
00:05:54,780 --> 00:06:00,900
Let's say feature on this code selection, underscore model.

82
00:06:00,910 --> 00:06:03,390
Let's say this is exactly its object.

83
00:06:03,600 --> 00:06:05,670
After it, I have to just execute it.

84
00:06:05,670 --> 00:06:10,030
Now, what we have to do, we have to basically fit my data to its object.

85
00:06:10,320 --> 00:06:12,660
So for this, I'm just going to say dot fit.

86
00:06:13,020 --> 00:06:20,870
And here I have to mention this Y and X, so I'm going to say X, comma Y, you just execute this and

87
00:06:20,880 --> 00:06:28,190
it will take a while and all the stuff in front of you, you will see your last model has been executed.

88
00:06:28,380 --> 00:06:34,590
And these are all the custom parameters that you have defined and these are all the things that you

89
00:06:34,590 --> 00:06:35,180
have done.

90
00:06:35,820 --> 00:06:37,350
So if you want to check.

91
00:06:37,470 --> 00:06:37,890
Yeah.

92
00:06:37,920 --> 00:06:42,600
What are the columns that are selected by these select from model class.

93
00:06:42,600 --> 00:06:50,550
So for this I'm just going to say feature on this selection model dot get on underscore support, which

94
00:06:50,550 --> 00:06:53,540
is exactly this one, and just execute this.

95
00:06:53,770 --> 00:07:00,670
You will see all your columns in the form of a list where you have values as central.

96
00:07:01,050 --> 00:07:07,020
So whenever there is a false, it means that model or you can say that feature selection technique is

97
00:07:07,020 --> 00:07:07,850
going to recommend.

98
00:07:09,040 --> 00:07:14,650
This feature isn't contributing more to my machine learning model and true means here.

99
00:07:14,880 --> 00:07:18,830
You have to consider this feature for your machine learning model.

100
00:07:19,290 --> 00:07:25,920
So for this, let's say I have to visualize here which feature gets elected and which feature doesn't

101
00:07:25,920 --> 00:07:26,730
get selected.

102
00:07:27,000 --> 00:07:31,740
So let's say I have to create a data frame or let's say you want to print it.

103
00:07:31,740 --> 00:07:34,320
Let's say whatever you want to have to print it, I have to print.

104
00:07:34,620 --> 00:07:39,650
So let's say if I'm going to call my expert columns, these are my all the columns.

105
00:07:39,660 --> 00:07:40,830
Let me store it somewhere else.

106
00:07:40,830 --> 00:07:44,200
Let's say these are exactly my all the columns over here.

107
00:07:44,490 --> 00:07:47,220
Now, what I have to do is I have to print something.

108
00:07:47,460 --> 00:07:51,360
So I'm going to say let's say I have to print what I have to print.

109
00:07:51,360 --> 00:07:56,970
I have to something like, say, my total features features.

110
00:07:56,970 --> 00:07:59,850
Then I have to add something known as a placeholder.

111
00:07:59,850 --> 00:08:05,550
So here I'm going to add my placeholder and this placeholder will receive value from my form and function.

112
00:08:05,550 --> 00:08:11,790
So I'm going to say a dot format and after it, what I have to do, I'm going to say X dot shape.

113
00:08:12,000 --> 00:08:14,190
And let's start with the number of column columns.

114
00:08:14,190 --> 00:08:19,260
So just pass one over there and in a similar way what you guys can do.

115
00:08:19,260 --> 00:08:25,710
Let me just copy this and let me just paste or did this time let's I have to print my select features.

116
00:08:25,740 --> 00:08:31,920
I'm going to say it is nothing but let's say my selected features and here what I have to pass.

117
00:08:32,160 --> 00:08:34,620
Let me just store it somewhere else.

118
00:08:34,620 --> 00:08:42,060
Let's say selected features are nothing, but let's say I have to class basically this filter in my

119
00:08:42,240 --> 00:08:42,810
column.

120
00:08:42,810 --> 00:08:48,240
So here I'm going to say I have to pass this filter in my column that said, if I'm going to execute

121
00:08:48,240 --> 00:08:53,610
it, you will get all your columns that have been selected from your select key model technique.

122
00:08:53,880 --> 00:08:54,770
So here you are.

123
00:08:54,780 --> 00:09:01,380
I have to do to store it somewhere else, I'm going to say selected on this score feature.

124
00:09:01,680 --> 00:09:03,780
So just execute it now.

125
00:09:03,780 --> 00:09:09,320
Here, what I have to do here, I am going to say I had to simply print it slant.

126
00:09:09,330 --> 00:09:14,400
I'm going to say basically land of whatever list I have defined.

127
00:09:14,410 --> 00:09:21,360
What and if I'm going to execute the salvator, you will see you have total features as trunked, but

128
00:09:21,360 --> 00:09:24,480
your model selects all your 14 features.

129
00:09:24,690 --> 00:09:30,710
So these are exactly your 14 features that you have to consider for the machine learning aspect.

130
00:09:31,140 --> 00:09:33,920
So let me just print what are exactly those features.

131
00:09:34,260 --> 00:09:36,990
So these are exactly all your features.

132
00:09:36,990 --> 00:09:40,290
It means you just need this feature to build a machine learning model.

133
00:09:40,290 --> 00:09:43,920
It means you have to consider those features as your independent features.

134
00:09:44,220 --> 00:09:48,660
So let me update my X, which is exactly my independent data frame.

135
00:09:48,660 --> 00:09:55,140
So I'm going to say if I'm going to pass this selected underscore feature in my X, so if I'm going

136
00:09:55,140 --> 00:09:58,590
to execute it now, you will see here you have this fourteen column.

137
00:09:58,680 --> 00:09:59,400
It mean this is.

138
00:09:59,470 --> 00:10:06,970
Your independent features, so let me just assign it an X that sit, so now you're dependent on data

139
00:10:06,970 --> 00:10:09,520
and independent data is almost ready.

140
00:10:09,520 --> 00:10:14,650
And in all of your upcoming sessions, we are going to build our model and we are also going to check

141
00:10:14,650 --> 00:10:21,520
what exactly the accuracy, how we can cross validate our model and all these things in a much more

142
00:10:21,520 --> 00:10:21,850
depth.

143
00:10:21,880 --> 00:10:24,390
So that's all about this session of the session.

144
00:10:24,430 --> 00:10:24,970
Very much.

145
00:10:25,270 --> 00:10:25,900
Thank you.

146
00:10:26,110 --> 00:10:27,040
Have a nice day.

147
00:10:27,340 --> 00:10:28,180
Keep learning.

148
00:10:28,180 --> 00:10:28,990
Keep growing.

149
00:10:29,110 --> 00:10:29,740
Keep practicing.