1
00:00:00,210 --> 00:00:06,390
Helen, in all our previous sessions, we have learned what is a disease and how you can build disease,

2
00:00:06,390 --> 00:00:12,510
and we have solved some of the use cases this last year that we have solved with respect to whether

3
00:00:12,510 --> 00:00:18,620
the person, whether the student of different class and different gender is going to stay on or not.

4
00:00:18,630 --> 00:00:23,640
We have solved this case using our our this approach.

5
00:00:23,670 --> 00:00:28,770
Similarly, we have some use case regarding our appropriate information approach.

6
00:00:28,770 --> 00:00:33,600
And we also learn what is pre brewing, what is postponing and indecision.

7
00:00:33,600 --> 00:00:37,680
We are going to learn what that and what it is and what are cons.

8
00:00:37,680 --> 00:00:43,820
And all of this is entry and what are different types of algorithms in decision.

9
00:00:43,840 --> 00:00:48,210
Very, very first, let's understand different algorithms.

10
00:00:48,230 --> 00:00:51,300
Let's understand our different algorithms.

11
00:00:52,350 --> 00:00:56,150
Let's understand our different algorithms for this, isn't there?

12
00:00:56,370 --> 00:01:00,680
There are literally tens of thousands of algorithms for this event.

13
00:01:01,170 --> 00:01:05,430
Let's say you have heard about it, which is eighty three a.m., which is nothing.

14
00:01:05,430 --> 00:01:08,330
My, my, I treated victimiser.

15
00:01:08,760 --> 00:01:16,130
So it is what the algorithm to construct a decision basically for our classification problem.

16
00:01:16,470 --> 00:01:19,200
It is basically for classification type of use cases.

17
00:01:19,360 --> 00:01:24,390
It is basically for classification used this Heidi, used by news algorithm.

18
00:01:25,080 --> 00:01:27,930
So how, how, how, how exactly.

19
00:01:28,070 --> 00:01:36,030
It creates diffidently said basically uses a concept of information gain as a criterion to build up

20
00:01:36,030 --> 00:01:36,540
this season.

21
00:01:36,930 --> 00:01:39,760
And whereas if I were to say so, I did.

22
00:01:39,930 --> 00:01:40,340
I did.

23
00:01:40,480 --> 00:01:47,760
He always works, always works in terms of in terms of categorical attribution.

24
00:01:47,850 --> 00:01:54,180
So whenever you have some categorical attribute, definitely go ahead with decided who internally,

25
00:01:54,450 --> 00:02:00,950
whenever you will see that you have some cycad learn Monu and whenever you have some let's say here

26
00:02:00,960 --> 00:02:08,830
you have a tree inside some module, you have something known as if you have two in this this tree.

27
00:02:09,080 --> 00:02:14,940
This isn't reclass decision tree either classifier or the question, depending upon what you guess you

28
00:02:14,940 --> 00:02:15,220
have.

29
00:02:15,570 --> 00:02:20,410
So let's say I'm going to import my classifier because I have a classification.

30
00:02:21,210 --> 00:02:28,920
So in this this season, tree classifier, if if we have this categorical attribute in our data, always

31
00:02:28,920 --> 00:02:34,350
go with this idea tree, you can you can customize all this barometer's if you if you will, if you

32
00:02:34,350 --> 00:02:39,470
will check all the different parameters in this in this class, you will find out you have you have

33
00:02:39,510 --> 00:02:41,490
tons of parameters inside this.

34
00:02:41,490 --> 00:02:42,930
And you can you play with that.

35
00:02:43,320 --> 00:02:43,710
Definitely.

36
00:02:43,800 --> 00:02:44,640
You can play with that.

37
00:02:45,660 --> 00:02:48,240
So now what what what we have to do list it.

38
00:02:48,900 --> 00:02:54,490
We have we have something known as we have something known as C 4.5.

39
00:02:55,440 --> 00:02:59,360
So as far as we know, it is it is just for categorical.

40
00:02:59,850 --> 00:03:00,720
But it is.

41
00:03:00,720 --> 00:03:03,330
It is it is for continuous.

42
00:03:03,960 --> 00:03:07,860
It is for continuous and for discrete data.

43
00:03:07,870 --> 00:03:12,990
It is especially for continuous and discrete type of data.

44
00:03:13,000 --> 00:03:14,080
So it is it is nothing.

45
00:03:14,100 --> 00:03:16,590
But it is just an extension.

46
00:03:16,590 --> 00:03:22,290
I can say extension of it is just an extension of this three and indefinitely.

47
00:03:22,400 --> 00:03:26,060
It is much, much better than I did much, much better than 93.

48
00:03:26,670 --> 00:03:31,140
It is definitely greater I against a much, much better than this idea here.

49
00:03:31,440 --> 00:03:35,660
And it always deals with this numerical data.

50
00:03:36,060 --> 00:03:42,700
It deals with numerical data as well as our categorical it is with both the scenarios, whether it's

51
00:03:42,720 --> 00:03:49,450
a numerical use case or against this regression, or I can say it's classification, but it is it is

52
00:03:49,530 --> 00:03:56,520
basically for classification and it basically uses our our Gini index.

53
00:03:56,550 --> 00:04:03,840
Or I guess you can import the guinea pig to select our parent or to select our group, because in this

54
00:04:03,960 --> 00:04:09,940
country the major concern is which in all is to be selected as a parent.

55
00:04:10,470 --> 00:04:17,790
So basically it it internally uses internally, it uses Kitney beauty to build, to build, to build

56
00:04:17,790 --> 00:04:18,650
up the season tree.

57
00:04:19,260 --> 00:04:21,900
So basically Padoan Cuttone.

58
00:04:22,970 --> 00:04:25,560
Carter highly was highly used against him.

59
00:04:25,910 --> 00:04:31,640
So basically a third one is exactly you have heard about something known as God, which is nothing.

60
00:04:31,640 --> 00:04:33,980
My classification.

61
00:04:34,130 --> 00:04:39,040
I can see classification and regression you have heard about.

62
00:04:39,260 --> 00:04:43,590
If you know a bit about this, you will definitely heard about this God.

63
00:04:44,150 --> 00:04:45,890
So how is God?

64
00:04:45,890 --> 00:04:54,320
Because this is basically so it also uses it also uses both in classification, both in classification

65
00:04:54,320 --> 00:04:55,700
as well as in regression.

66
00:04:55,790 --> 00:05:02,290
You definitely you can go with this card because it's a highly used algorithm in highly used.

67
00:05:03,450 --> 00:05:13,500
So how how exactly it builds isn't internally, it uses something known as Guinea Index as as as I guess

68
00:05:13,500 --> 00:05:21,900
as a barometer, as a default barometer to select the select rulebook as as how how I did select how

69
00:05:22,080 --> 00:05:30,480
how my seat 4.5 five and select our our this type of approach, which is my guinea border post to select

70
00:05:30,480 --> 00:05:30,990
Google.

71
00:05:31,410 --> 00:05:36,810
But if you want to do modification, if you want to do modification in this, if you want to do some

72
00:05:36,810 --> 00:05:39,240
kind of modification, definitely we can do that.

73
00:05:39,240 --> 00:05:46,290
We can use Entropia as well in this in this in this card we can use we can definitely entropies definitely

74
00:05:46,290 --> 00:05:47,460
we can definitely use this.

75
00:05:48,330 --> 00:05:52,470
Similarly, we have something known as five point zero.

76
00:05:52,650 --> 00:05:54,410
We have something see 5.0.

77
00:05:54,720 --> 00:06:00,480
So it is definitely it is it is better than SI four point five, much, much better than this.

78
00:06:01,080 --> 00:06:06,480
And it having however it has its own content does it.

79
00:06:06,480 --> 00:06:08,880
It works with categorical data.

80
00:06:08,880 --> 00:06:10,470
It works with categorical.

81
00:06:11,470 --> 00:06:19,650
Target, I can say it works categorically, target, whenever your target variable or I can see whenever

82
00:06:19,670 --> 00:06:26,140
level is of categorical feature, let's say yes no like that in that type of scenario.

83
00:06:26,180 --> 00:06:33,130
This this highly used to the similarly we have something known as Musse that is not used that much,

84
00:06:33,400 --> 00:06:41,890
rarely used Gore indeed in this decision tree, which is my thing, but multi variant adaptive regression

85
00:06:42,100 --> 00:06:43,330
Desplaines I can see.

86
00:06:43,480 --> 00:06:45,510
So what, what actually does.

87
00:06:46,150 --> 00:06:55,140
So it always try to create a series of beer with linear model beer with a linear model.

88
00:06:55,150 --> 00:07:02,950
Basically it it creates a series of pairwise, pairwise, pairwise in a model which is being used,

89
00:07:03,340 --> 00:07:09,500
which is being used to modulate irregularity and interactions among all variables.

90
00:07:09,640 --> 00:07:13,900
And this is not a widely used algorithm that's not highly used algorithm.

91
00:07:14,470 --> 00:07:16,000
And this isn't yet.

92
00:07:16,000 --> 00:07:21,880
But yeah, this guy and I did and see 4.5 definitely highly used.

93
00:07:21,880 --> 00:07:29,860
And most other time, as in the Askia, what type of algorithm, you know, when you are going to build

94
00:07:30,610 --> 00:07:36,280
a facility, will ask you whether you know this for twenty four point five, whether you know this card

95
00:07:36,280 --> 00:07:40,150
or whether you know, IDC, what type of algorithm.

96
00:07:40,180 --> 00:07:41,290
And we have many more.

97
00:07:41,560 --> 00:07:43,180
Many more, many more.

98
00:07:43,180 --> 00:07:50,680
Like we have something known as or this is a system we have something known as this isn't as dumb as

99
00:07:50,680 --> 00:07:50,930
that.

100
00:07:50,950 --> 00:07:52,790
So it is nothing.

101
00:07:52,810 --> 00:07:54,780
But it is just one label.

102
00:07:54,790 --> 00:07:55,510
This is it.

103
00:07:55,560 --> 00:08:00,280
And it is used for generating a decision tree by just a single.

104
00:08:00,610 --> 00:08:02,800
We have to just do a single spit.

105
00:08:03,460 --> 00:08:03,940
That's it.

106
00:08:03,940 --> 00:08:05,200
That that's all about the.

107
00:08:05,470 --> 00:08:08,730
And we have something known as something on it.

108
00:08:09,430 --> 00:08:11,310
So what, how this works.

109
00:08:11,530 --> 00:08:18,610
So whenever, whenever you have you have much analyzing data or I can say you have outliers in data

110
00:08:18,610 --> 00:08:19,630
whenever you have some.

111
00:08:19,870 --> 00:08:24,300
I can say that say you have some small little X that you have some small dataset.

112
00:08:24,340 --> 00:08:27,970
Yeah, I have I have worked with I have worked with this algorithm.

113
00:08:27,970 --> 00:08:29,470
I work with this type of algorithm.

114
00:08:29,590 --> 00:08:29,930
Yeah.

115
00:08:30,310 --> 00:08:36,480
So whenever you have some small data and let's say that, that that data has outliers and that has some

116
00:08:36,490 --> 00:08:41,530
noise kind of thing, go ahead with this and five tables and just go ahead.

117
00:08:41,530 --> 00:08:49,000
Would definitely you seem a darn good accuracy when you go ahead with this and 5000 of this of this

118
00:08:49,000 --> 00:08:49,540
algorithm.

119
00:08:49,690 --> 00:08:56,380
And we have something known as is there are lots of things like this, nothing but just my conditional

120
00:08:56,380 --> 00:08:59,720
investigations, all sorts of Elgort overbuilding, incidentally.

121
00:09:00,430 --> 00:09:01,100
But but yeah.

122
00:09:01,150 --> 00:09:04,660
But at least you should know, what is this ADT?

123
00:09:05,080 --> 00:09:06,880
What are my card algorithms?

124
00:09:06,880 --> 00:09:10,500
And, you know, what is the C point C for five.

125
00:09:10,510 --> 00:09:17,040
And and if if you are OK, then you can consider this and five and see 5.5.

126
00:09:17,050 --> 00:09:22,570
I will, I will recommend you at least at least you should know this to at least at least you should

127
00:09:22,580 --> 00:09:22,700
know.

128
00:09:22,930 --> 00:09:26,610
And if you are OK with this, definitely you can consider this too as well.

129
00:09:26,710 --> 00:09:31,990
But but actually try to learn this to try to learn because if if if you really want to.

130
00:09:31,990 --> 00:09:32,360
Good.

131
00:09:32,440 --> 00:09:37,510
If you really want to go into into machine learning, if you really want to go into machine learning,

132
00:09:37,520 --> 00:09:42,850
do definitely you can you can see that all this type of thing, all this fight at least at least you

133
00:09:42,850 --> 00:09:44,380
have to go through this fight if you want to.

134
00:09:44,500 --> 00:09:50,260
If you really want to go with machine learning and you have something known as this security, which

135
00:09:50,260 --> 00:09:56,530
is nothing which which works some dumb thing on a statistical test with something on Chi Square test,

136
00:09:57,820 --> 00:10:05,440
which is nothing like this, might chi square automatic interaction detector dexa of algorithm for to

137
00:10:05,460 --> 00:10:05,650
build.

138
00:10:06,560 --> 00:10:10,070
So let's talk about what are the consent forms of the season.

139
00:10:10,520 --> 00:10:12,980
Let's talk about, let's talk about.

140
00:10:13,450 --> 00:10:18,210
So let's, let's talk about what very fourth quarter and one day this or this isn't.

141
00:10:18,700 --> 00:10:24,610
So the very first is it is used both in case of regression as well as in case of classification.

142
00:10:25,030 --> 00:10:32,560
That's the very first one is definitely and the very first and very second one is exactly very easy

143
00:10:32,560 --> 00:10:33,040
to learn.

144
00:10:33,040 --> 00:10:33,970
It's very easy.

145
00:10:33,970 --> 00:10:41,410
It's very damn easy to learn because nothing you can just done buton Loades, just entropic type of

146
00:10:41,410 --> 00:10:44,130
things and you have to learn something.

147
00:10:44,150 --> 00:10:50,650
One is getting Gote, but just you have to learn to two things, just just two things to learn it.

148
00:10:51,040 --> 00:10:57,790
And the rules of splitting is clearly mentioned as as as we have learned how to create this.

149
00:10:57,790 --> 00:11:03,790
Isn't we using you know, we have learned this using this skinny necks, this entropy information gain

150
00:11:03,790 --> 00:11:06,340
like I never told anyone tases.

151
00:11:06,340 --> 00:11:07,240
Exactly.

152
00:11:07,330 --> 00:11:09,670
We don't have to deal with outliers.

153
00:11:09,670 --> 00:11:10,960
We don't have to deal with.

154
00:11:11,670 --> 00:11:20,460
Outliers in case of this season, because this isn't is exactly what's on the basis of this kind of

155
00:11:20,480 --> 00:11:21,330
rule-based thing.

156
00:11:21,690 --> 00:11:26,130
So whenever you have outlier, it doesn't get impacted that much.

157
00:11:27,090 --> 00:11:31,270
Similarly, similarly, you have something known as is killing.

158
00:11:31,530 --> 00:11:39,600
You have heard about this is killing and something of this normalization, something known as normalization.

159
00:11:39,840 --> 00:11:41,700
So there is no need to perform.

160
00:11:41,700 --> 00:11:43,920
This is scaling and normalization.

161
00:11:43,930 --> 00:11:45,450
There is no need to perform.

162
00:11:45,900 --> 00:11:52,350
This is scaling and normalization in case of this disease and in case of disease.

163
00:11:52,350 --> 00:11:57,900
And there is no need to perform this as as this disease and treat as a disease.

164
00:11:57,900 --> 00:12:05,430
And three entirely works on the basis of the season that greater than these whether it's less than these

165
00:12:05,440 --> 00:12:13,200
is so there is no need to do is killing and not for that are exactly my all the advantages of this.

166
00:12:14,130 --> 00:12:16,560
Let's talk about let's talk about water.

167
00:12:16,950 --> 00:12:19,030
What are some this I don't want this.

168
00:12:19,050 --> 00:12:20,550
What are some discussion on this?

169
00:12:21,150 --> 00:12:28,830
Let's talk about this for the very first tell that I have filled out in my walking with this algorithm

170
00:12:28,830 --> 00:12:30,750
in something some real world scenarios.

171
00:12:31,470 --> 00:12:37,050
So what if we are going to make let's say, if we are going to make a small change in data, let's say

172
00:12:37,050 --> 00:12:42,300
we are going to make some small change in let's say, in creating data, not in data.

173
00:12:42,330 --> 00:12:43,670
I can see it as training.

174
00:12:43,680 --> 00:12:51,720
You do so whenever we are going to make a small change or a minor change in this training data, what

175
00:12:51,720 --> 00:12:54,540
it will do, it will cause on a stability.

176
00:12:54,690 --> 00:13:00,120
Or if we make a small change in training data, basically it will shift the node.

177
00:13:00,420 --> 00:13:03,420
It will shift toward, let's say this is DLT.

178
00:13:03,870 --> 00:13:09,420
So if we do a small change, if we do a small change, it will let's say this is my entire decision.

179
00:13:09,420 --> 00:13:12,840
Let's say this is my let's that this is my own time.

180
00:13:13,440 --> 00:13:17,850
So if we make a small change, let's see over here, I have made a small change.

181
00:13:18,090 --> 00:13:24,750
So it will make a shift of it will make a shift of this this node, this node from one side to the other

182
00:13:24,750 --> 00:13:25,140
side.

183
00:13:25,290 --> 00:13:32,820
And many other reasons that this is not a disadvantage that I have faced, that I have faced when I

184
00:13:32,850 --> 00:13:34,770
was working in some real world scenarios.

185
00:13:35,220 --> 00:13:44,460
And definitely one more advantage is definitely a problem of all this is very, very high, is very,

186
00:13:44,460 --> 00:13:49,890
very high, is is very, very high, literally very, very high in case of encephalopathies tree.

187
00:13:50,160 --> 00:13:56,910
So we have something known as pruning tool to to get rid of this autophagy and tired of it and Tadamon.

188
00:13:56,910 --> 00:13:57,450
Exactly.

189
00:13:58,080 --> 00:14:00,720
It takes more time to train and this isn't a model.

190
00:14:00,840 --> 00:14:05,040
Definitely takes more time because it takes more time.

191
00:14:05,940 --> 00:14:09,570
It takes more time to Crenna.

192
00:14:11,140 --> 00:14:17,470
This season, Trimodal, it definitely takes more time to clean and remodel than under classification

193
00:14:17,470 --> 00:14:25,180
algorithms like like autistic and many others, because as we do a lot of spreading, we have one here.

194
00:14:25,180 --> 00:14:28,780
Obviously, we have lots of earsplitting or here or here or here.

195
00:14:29,030 --> 00:14:35,180
We have lots of it spreading out here and lots of calculation if if I would consider entropy.

196
00:14:36,070 --> 00:14:40,030
So we have to consider entropy and let's say information again.

197
00:14:40,250 --> 00:14:45,220
Then we have to do lots of also splitting and we will end up having the misandry.

198
00:14:46,060 --> 00:14:48,870
That's why that's why it will it will take more than anything.

199
00:14:49,690 --> 00:14:52,660
So that's all about the session holtsville of the session?

200
00:14:52,670 --> 00:14:53,650
Very much so.

201
00:14:53,650 --> 00:14:59,080
In the upcoming session, we are going to learn how this essentially works, actually, in case of regression,

202
00:14:59,080 --> 00:15:03,730
because we have learned how it exactly works in case of classification.

203
00:15:03,940 --> 00:15:09,580
Whenever we have some discrete data and whenever we have a continuous data show in the upcoming session,

204
00:15:09,580 --> 00:15:14,650
we are going to learn how how actually it works in case of some some regression you.

205
00:15:15,430 --> 00:15:17,190
So we have to understand as well this.

206
00:15:17,650 --> 00:15:18,800
So that's all about it.

207
00:15:18,980 --> 00:15:20,530
I hope you will love this very much.

208
00:15:20,800 --> 00:15:21,350
Thank you.

209
00:15:21,400 --> 00:15:22,270
Have a nice day.

210
00:15:22,810 --> 00:15:23,680
Keep learning.

211
00:15:23,680 --> 00:15:24,580
Keep growing.

212
00:15:24,850 --> 00:15:25,690
Keep practicing.