1
00:00:00,470 --> 00:00:06,810
All so in the previous session, what we have done, we have basically extract some desired features

2
00:00:06,810 --> 00:00:12,720
from our data and then all of our sessions in all our previous sessions, we have analyzed our data

3
00:00:12,720 --> 00:00:14,550
as well as we have understand our data.

4
00:00:14,940 --> 00:00:18,640
And that's a time to approach for your machine learning.

5
00:00:18,930 --> 00:00:25,370
So in this session, what we are going to do, we have this amazing assignment in which we have to apply

6
00:00:25,770 --> 00:00:28,490
feature encoding techniques on data.

7
00:00:28,500 --> 00:00:34,110
So very first, let's understand why there is a need of feature encoding techniques and what are different,

8
00:00:34,140 --> 00:00:36,240
different type of encoding techniques.

9
00:00:36,820 --> 00:00:42,240
So, yeah, so you will figure out here you have direct, here you have direct and all these kinds of

10
00:00:42,330 --> 00:00:46,120
string like things like machine learning now understand an extension data.

11
00:00:46,140 --> 00:00:53,220
So it means we have to convert this data and do some numerical format using some feature encoding techniques.

12
00:00:53,400 --> 00:00:57,880
That's why feature encoding techniques come into existence.

13
00:00:58,110 --> 00:01:05,160
So in this session, what we are going to do, we are basically going to use our technique, a very

14
00:01:05,160 --> 00:01:09,650
popular technique known as mean encoding technique.

15
00:01:09,900 --> 00:01:14,320
So let's understand what exactly the mean encoding technique.

16
00:01:14,790 --> 00:01:20,750
So if you will figure out in this market segment, let's say in this market segment, here you have

17
00:01:21,000 --> 00:01:22,500
so let me show you a thing.

18
00:01:22,590 --> 00:01:28,710
Data underscore, get of market segment, just press tab.

19
00:01:28,830 --> 00:01:35,220
And if on this I'm going to call my unique and if I'm just going to execute it, you will see it has

20
00:01:35,220 --> 00:01:37,710
that much unique number of categories.

21
00:01:37,860 --> 00:01:45,870
So what meaning coding will do mean encoding will basically see whatever will be the mean with respect

22
00:01:45,870 --> 00:01:47,670
to this direct column.

23
00:01:47,670 --> 00:01:55,110
Or you can say whatever the mean of the direct with respect to this cancellation column, because cancellation

24
00:01:55,110 --> 00:01:58,290
is exactly my independent feature that we have to predict.

25
00:01:58,320 --> 00:02:03,170
So what are what will be the nien of this act with respect to this cancellation?

26
00:02:03,610 --> 00:02:06,830
I'm basically going to replace Garecht with that mean.

27
00:02:06,840 --> 00:02:09,240
That's what my mean encoding will do.

28
00:02:09,360 --> 00:02:14,360
So if you're not that much comfort, let me show you a code so that you all can understand.

29
00:02:14,550 --> 00:02:20,400
So here what we have to do very first, let's say we have to extract all those columns on which we have

30
00:02:20,400 --> 00:02:21,590
to perform these techniques.

31
00:02:22,050 --> 00:02:28,860
So if on this data underscore that if I'm going to call my columns or they are just execute and these

32
00:02:28,860 --> 00:02:37,080
are exactly all the columns, let's say here, if I'm going to say in these columns, I just don't need

33
00:02:37,080 --> 00:02:42,330
this cancellation because I don't have to encode this feature because this is my defining feature and

34
00:02:42,330 --> 00:02:44,620
this is that what we have to predict.

35
00:02:44,940 --> 00:02:48,480
So it means I'm going to cost zero to it.

36
00:02:48,630 --> 00:02:50,610
And here I'm going to say zero to tailgated.

37
00:02:50,820 --> 00:02:55,200
And if I'm going to execute it, you will see you don't have this cancellation column.

38
00:02:55,230 --> 00:03:01,710
Now I'm going to store it somewhere else, let's say columns, and I'm just going to print it as columns

39
00:03:01,710 --> 00:03:02,640
just execute.

40
00:03:02,760 --> 00:03:07,100
And you will figure out these are my all those columns that we have to include.

41
00:03:07,110 --> 00:03:13,920
So I'm going to say for column in this cost list, it means I have to think it means I have to fetch

42
00:03:13,920 --> 00:03:14,840
each and every column.

43
00:03:15,300 --> 00:03:19,010
Then I'm going to see data on the scorecard.

44
00:03:19,020 --> 00:03:24,350
And this time I have to group my data on the basis of each and every column for this.

45
00:03:24,360 --> 00:03:29,610
I'm just going to say, let me show you a very, very basic example over here.

46
00:03:29,850 --> 00:03:36,380
Let's say I'm going to say data on a scorecard and let's say I'm just going to access my hotel.

47
00:03:36,630 --> 00:03:39,900
So if I'm going to exit this hotel, you will see all the stats.

48
00:03:40,380 --> 00:03:45,270
And now what I have to do if I'm with the group on the basis of this hotel.

49
00:03:45,300 --> 00:03:49,770
So for this, I'm going to say I have to group on the basis of all this hotel.

50
00:03:50,310 --> 00:03:58,290
Once I have this group, by then I'm going to access my calculation, which is exactly this one.

51
00:03:58,300 --> 00:04:00,410
So I have to just access this.

52
00:04:00,780 --> 00:04:04,800
So I'm going to say this is exactly my calculation.

53
00:04:04,800 --> 00:04:09,660
And on this, if I am going to call me Nordea, so let me just call me.

54
00:04:09,690 --> 00:04:16,410
And if I'm going to execute, you will see with respect to City Hall Hotel, you have that much me with

55
00:04:16,410 --> 00:04:20,090
respect to this resort hotel, you have that much mean.

56
00:04:20,370 --> 00:04:27,960
It means what you can do wherever you have a city hotel, just replace it with zero one, two seven.

57
00:04:28,140 --> 00:04:33,360
And internally, my machine learning model will able to understand what exactly the meaning of the zero

58
00:04:33,360 --> 00:04:37,740
point two seven, because machine learning now understand what a city hotel is.

59
00:04:37,760 --> 00:04:40,800
Just understand my Intisar and data.

60
00:04:40,810 --> 00:04:42,950
So we have to tell the machine learning model.

61
00:04:42,970 --> 00:04:45,430
Yeah, this is exactly my city hotel.

62
00:04:45,450 --> 00:04:47,580
So this is that approach that you have to follow.

63
00:04:47,580 --> 00:04:50,010
And this is exactly what I mean.

64
00:04:50,010 --> 00:04:56,010
In quoting now, you have to perform, you have to perform this approach for each and every feature.

65
00:04:56,040 --> 00:04:59,810
So here I am going to say I have to just perform this at.

66
00:05:00,200 --> 00:05:06,800
For each and every feature here, I have to say, this gets replaced by basically my column, that's

67
00:05:06,800 --> 00:05:07,050
it.

68
00:05:07,490 --> 00:05:12,670
Now, what I have to do, basically, let's say I have to print something, Sarah.

69
00:05:12,710 --> 00:05:14,600
I'm going to say just print it.

70
00:05:15,200 --> 00:05:21,560
And let's say after printing to make it more user friendly, you can add some spaces as well.

71
00:05:21,670 --> 00:05:26,990
I'm just going to add and just going to execute, you will see with respect to all that, you have that

72
00:05:26,990 --> 00:05:31,940
much value with respect to resort hotel, you have that much value and all this sort of thing.

73
00:05:32,390 --> 00:05:39,350
Now, what you have to do basically here, you have to simply convert all these stuffs into a dictionary.

74
00:05:39,390 --> 00:05:43,120
Now, the question that you guys can ask or is a query.

75
00:05:43,610 --> 00:05:43,850
Yeah.

76
00:05:43,850 --> 00:05:45,940
Why why there is a need to convert this.

77
00:05:45,950 --> 00:05:51,350
All this is just introduction because you have to map you that our city hotel.

78
00:05:51,350 --> 00:05:53,420
I have to map this value.

79
00:05:53,660 --> 00:05:57,820
And once you will create a dictionary now data is in the form of key value pairs.

80
00:05:57,830 --> 00:06:02,650
It means the sea city hotels become key and this becomes value.

81
00:06:03,080 --> 00:06:08,000
So once you have all this stuff in your dictionary, you can easily map that.

82
00:06:08,450 --> 00:06:09,890
That's the power of dictionary.

83
00:06:10,250 --> 00:06:15,250
So here I am going to say I have to convert all this stuff into the city.

84
00:06:15,260 --> 00:06:17,500
Just called to underscore DECT over there.

85
00:06:17,510 --> 00:06:19,310
And again, I'm going to execute it now.

86
00:06:19,310 --> 00:06:23,550
You will see you have all these stuffs available over here.

87
00:06:23,690 --> 00:06:26,050
All this is now simple.

88
00:06:26,060 --> 00:06:26,770
It's simple.

89
00:06:26,780 --> 00:06:29,800
Now you have to just you have to just map it.

90
00:06:30,260 --> 00:06:32,200
So far, this is what I'm going to do.

91
00:06:32,210 --> 00:06:35,570
Let's say this time I don't have to pay Latza this time.

92
00:06:35,570 --> 00:06:40,580
This is exactly my dictionary with respect to each and every column.

93
00:06:40,940 --> 00:06:43,550
So here I going to say this is exactly my dictionary.

94
00:06:44,000 --> 00:06:48,380
Once I have all this stuff, what I have to do, I have to just map it.

95
00:06:48,380 --> 00:06:54,470
So for this, I'm going to say data on the score, get off column, dot the map.

96
00:06:54,470 --> 00:06:57,200
I have to map what my dictionary is simple.

97
00:06:57,620 --> 00:07:00,430
Then I have to update my net column as well.

98
00:07:00,650 --> 00:07:07,270
So data and a score get of column equals to this one index.

99
00:07:07,280 --> 00:07:08,960
It just executed.

100
00:07:08,960 --> 00:07:09,950
It will take a time.

101
00:07:09,950 --> 00:07:19,490
And again, if I'm going to call my head over there now, you will figure out all your string data gets

102
00:07:19,490 --> 00:07:26,600
converted into some the values and your machine learning model now able to understand internally what

103
00:07:26,600 --> 00:07:31,370
exactly the meaning of this, because machine learning how to understand is Janita.

104
00:07:32,060 --> 00:07:38,390
So if you guys are not that much comfort level about this, meaning coding, you can definitely raise

105
00:07:38,390 --> 00:07:43,180
your credit in your section or you can personally text me as well.

106
00:07:43,730 --> 00:07:46,030
So I hope you will love this approach as well.

107
00:07:46,040 --> 00:07:50,690
There are tons of approaches to convert this into something into the format, like the label encoding

108
00:07:50,690 --> 00:07:51,500
and all these things.

109
00:07:51,890 --> 00:07:56,510
But here I'm going to show you some advance support, because whenever you are going to work on some

110
00:07:56,510 --> 00:08:00,790
real work data there, you have to use some advanced approach.

111
00:08:00,800 --> 00:08:04,650
You don't have to use some approach like dummy approaches.

112
00:08:04,700 --> 00:08:11,060
Now, what you have to do, you need your entire data frame in which you have all the numerical features

113
00:08:11,060 --> 00:08:13,010
in which we have all the categorical features.

114
00:08:13,510 --> 00:08:17,540
So for this, what I'm going to do, I'm just going to concatenate my two data.

115
00:08:18,140 --> 00:08:26,150
So here I'm going to say in this data underscore get and after it, I have something.

116
00:08:26,330 --> 00:08:31,580
So here I am going to say data of num underscore features.

117
00:08:31,760 --> 00:08:38,390
After having all this stuff, I have to mention something my access equals to one, because here I'm

118
00:08:38,390 --> 00:08:43,860
going to say I have to concatenate in vertical fashion after doing all this stuff.

119
00:08:43,880 --> 00:08:45,350
Let me summarize.

120
00:08:45,350 --> 00:08:48,140
Let's say this is my entire data.

121
00:08:48,440 --> 00:08:50,170
I'm going to say this is own data frame.

122
00:08:50,210 --> 00:08:51,630
Just execute this cell.

123
00:08:51,950 --> 00:09:00,050
And if I'm going to call this data frame, Dot had to get a quick overview of how my data frame looks

124
00:09:00,050 --> 00:09:00,340
like.

125
00:09:00,380 --> 00:09:04,970
Now, you will see these are exactly all the features that I have over here.

126
00:09:05,000 --> 00:09:11,260
And you will see over here there are two features that are going to repeat it, which are exactly this

127
00:09:11,270 --> 00:09:15,780
cancellation and cancel it means you can drop any one of them.

128
00:09:16,100 --> 00:09:21,580
So for this, I'm just going to say data frame, dot drop.

129
00:09:21,590 --> 00:09:23,380
I have to drop what I have to do up.

130
00:09:23,390 --> 00:09:28,950
I have to drop this feature the symbol then I have to access equals to one because I have to remove

131
00:09:29,150 --> 00:09:30,260
in a vertical way.

132
00:09:30,260 --> 00:09:32,700
Then I'm going to say I have to update it as well.

133
00:09:33,020 --> 00:09:35,920
So in place equals to true.

134
00:09:36,590 --> 00:09:39,530
Just execute all this just gets executed.

135
00:09:39,530 --> 00:09:46,580
And if I'm going to call my data or shape forward here now, you will see it has that much number of

136
00:09:46,580 --> 00:09:49,910
rows and it has that much number of columns.

137
00:09:49,920 --> 00:09:51,580
So that's all about the session.

138
00:09:51,650 --> 00:09:58,130
In the upcoming session, we are going to decide what exactly the outline, how to handle outlier situations

139
00:09:58,130 --> 00:09:58,850
in your data.

140
00:09:59,330 --> 00:09:59,660
So who.

141
00:10:00,020 --> 00:10:06,680
Discussion very much, and you still have a query you can feel free to ask, just raise in your section

142
00:10:06,710 --> 00:10:09,160
or you can personally text me as well.

143
00:10:09,170 --> 00:10:10,260
So thank you.

144
00:10:10,300 --> 00:10:11,170
Have a nice day.

145
00:10:11,350 --> 00:10:12,260
Keep learning.

146
00:10:12,260 --> 00:10:13,160
Keep growing.

147
00:10:13,340 --> 00:10:14,210
Keep practicing.