1
00:00:00,450 --> 00:00:05,560
Hello, guys, and welcome back to the class, of course, for the complete introduction to data science.

2
00:00:06,120 --> 00:00:10,050
So in this class, we are going to talk about machine learning.

3
00:00:10,440 --> 00:00:16,230
So in the past class, we talk about what is it in general, how it works and basically all the basics

4
00:00:16,230 --> 00:00:16,740
of this.

5
00:00:17,160 --> 00:00:21,750
In this class, we are going to talk about the different machine learning algorithms.

6
00:00:21,960 --> 00:00:25,770
So we're not going to go into practice of those algorithms.

7
00:00:26,250 --> 00:00:29,180
We are going to cover all the theoretical part.

8
00:00:29,460 --> 00:00:32,020
So I'm going to talk about each of those.

9
00:00:32,040 --> 00:00:39,180
Well, each of the of the most important machine learning algorithms, how they work and simply give

10
00:00:39,180 --> 00:00:43,880
you a brief introduction to each of those algorithms.

11
00:00:43,920 --> 00:00:49,920
Once again, my goal is not that you guys become professional in all this, but simply that you understand

12
00:00:49,920 --> 00:00:57,120
what is it how each of those algorithms works and basically all that you can do with machine learning

13
00:00:57,120 --> 00:00:58,410
and Python as well.

14
00:00:58,770 --> 00:00:59,750
So it's not waiting more.

15
00:00:59,790 --> 00:01:01,560
And let's jump right into it.

16
00:01:01,860 --> 00:01:02,290
All right.

17
00:01:02,820 --> 00:01:08,340
So we see the first machine learning algorithm that we are going to talk about will be linear regression.

18
00:01:08,760 --> 00:01:12,060
And this one is pretty simple to understand.

19
00:01:12,660 --> 00:01:18,480
So basically, the formula is Y equal to X plus B. So basically, as you can see on the graph right

20
00:01:18,480 --> 00:01:18,850
here.

21
00:01:19,710 --> 00:01:23,080
So what exactly is the linear regression in machine learning?

22
00:01:23,850 --> 00:01:31,890
It's simply a supervised machine learning algorithm where the prediction output is continuous and has

23
00:01:31,920 --> 00:01:36,330
a well, we can say concrete sloup.

24
00:01:36,690 --> 00:01:40,890
Basically the slope is pretty much the same since it's Y equals X plus.

25
00:01:42,000 --> 00:01:47,100
So basically an example of machine learning regression could be the graph right here.

26
00:01:47,110 --> 00:01:53,970
So we can say, for example, that right here we have the price paid for, for example, let's say a

27
00:01:53,970 --> 00:01:54,420
plate.

28
00:01:54,430 --> 00:02:00,630
So let's say in the restaurant you guys are buying something and this would be the price paid for your

29
00:02:00,630 --> 00:02:02,370
dish or for what you have paid.

30
00:02:02,850 --> 00:02:07,830
And this would be the amount of thip of that you guys have left to basically, as you can see, this

31
00:02:07,830 --> 00:02:09,830
would be our well, our test.

32
00:02:10,290 --> 00:02:17,190
So the system will be able to tell us that the more someone buys the well, the more the plate is expensive,

33
00:02:18,090 --> 00:02:20,210
the more the person will leave a tip.

34
00:02:20,220 --> 00:02:21,930
So the more the tip will be high.

35
00:02:22,890 --> 00:02:27,540
So basically, as you can see, and there would be right here, we will have a in the middle.

36
00:02:27,930 --> 00:02:30,510
We'll have our food that will be right there.

37
00:02:30,520 --> 00:02:33,550
So you can see it's it's drawing right here.

38
00:02:33,570 --> 00:02:36,450
So basically, this would be like the average or the mean.

39
00:02:37,140 --> 00:02:39,870
Basically, it's in the middle of all the points.

40
00:02:40,510 --> 00:02:43,960
Some linear regressions are, well, better than others.

41
00:02:43,980 --> 00:02:46,170
So in some of them, you will have points everywhere.

42
00:02:46,710 --> 00:02:50,700
And some of them, the points are very, very well condensed in one place.

43
00:02:51,660 --> 00:02:53,350
Those are more representative.

44
00:02:53,370 --> 00:02:58,500
So basically a soup like this would be more representative that if we have points everywhere and there

45
00:02:58,500 --> 00:03:01,860
is one soup in the average, it's not really representative.

46
00:03:02,850 --> 00:03:09,480
So this is for linear regression, the same type of algorithms that we are going to be the logistic

47
00:03:09,840 --> 00:03:10,530
regression.

48
00:03:11,160 --> 00:03:13,890
And it's not like linear regression.

49
00:03:13,920 --> 00:03:16,820
Basically, this is a logistic regression.

50
00:03:17,910 --> 00:03:23,340
Well, it comes from statistics like the linear regression, but it's not that long ago riddim for regression

51
00:03:23,340 --> 00:03:23,850
problems.

52
00:03:24,540 --> 00:03:30,870
In other words, problems that require the prediction of a continuous outcome.

53
00:03:31,350 --> 00:03:38,220
So basically, it's not for this absolute logistic regression is really a method that is used for binary

54
00:03:38,220 --> 00:03:39,180
classifications.

55
00:03:39,180 --> 00:03:41,520
In other words, one or zero.

56
00:03:42,210 --> 00:03:45,180
So it will give you a discrete binary outcome between zero.

57
00:03:45,180 --> 00:03:46,120
And so, as I said.

58
00:03:47,340 --> 00:03:53,310
So if we want to think about something so basically we can have well, as an example, we can, for

59
00:03:53,310 --> 00:03:58,910
example, say, do you like to eat outside to basically asking someone that if he likes to eat outside,

60
00:03:58,920 --> 00:04:03,210
yes or no, I could be a logistic regression and zero in the square?

61
00:04:03,260 --> 00:04:09,590
Well, yes, let's say yes in this case will be zero and no in this case would be one.

62
00:04:10,290 --> 00:04:12,180
So basically it will look something like this.

63
00:04:12,180 --> 00:04:15,990
So you can see all will have zero right here and we'll have one right here.

64
00:04:16,290 --> 00:04:20,270
And after that, we will have an average and this would be our outcome.

65
00:04:20,280 --> 00:04:26,970
So we see the majority of people like to eat outside if it's closer to one, because one is yes and

66
00:04:26,970 --> 00:04:27,840
zero is no.

67
00:04:28,290 --> 00:04:33,630
So basically, this is another type of regression that we can use in machine learning.

68
00:04:33,630 --> 00:04:37,530
But once again, this one is really, really important, like the first one.

69
00:04:37,530 --> 00:04:43,860
And the difference between linear and logistic is that logistic is really between zero and one.

70
00:04:43,870 --> 00:04:48,210
So this is more of a binary.

71
00:04:48,220 --> 00:04:55,620
Well, the outcome that you will have will be a discrete binary outcome, not like the linear regression

72
00:04:55,950 --> 00:04:59,910
to the algorithm that we are going to talk about in this class is pretty.

73
00:05:00,340 --> 00:05:02,190
It's called the decision tree.

74
00:05:02,650 --> 00:05:09,080
So I think we already talked about this type of algorithm at the beginning of the course basically about

75
00:05:09,100 --> 00:05:11,590
decision trees and other stuff related to this.

76
00:05:12,640 --> 00:05:15,310
So basically, what exactly is a decision tree?

77
00:05:16,180 --> 00:05:23,950
So basically, decision trees well, in general are constructed via an algorithm approach that identifies

78
00:05:23,950 --> 00:05:27,340
ways to split a dataset based on different conditions.

79
00:05:28,150 --> 00:05:32,710
So basically, we will have different conditions for different variables that we will need to take input

80
00:05:32,710 --> 00:05:37,010
into consideration to be able to create our decision tree.

81
00:05:38,050 --> 00:05:45,430
So it's once again one of the most used and practical methods because it's really based on how we as

82
00:05:45,430 --> 00:05:47,330
human beings solve our problems.

83
00:05:47,740 --> 00:05:50,680
So basically, let's take an example.

84
00:05:50,680 --> 00:05:52,610
So the example right here that we have.

85
00:05:53,140 --> 00:05:58,150
So once again, this is a decision tree that will help us know what animal we are talking about.

86
00:05:58,570 --> 00:06:03,340
So basically, the first question that we are going to ask is, does he has feeder's if yes.

87
00:06:03,340 --> 00:06:04,090
Can he fly?

88
00:06:04,210 --> 00:06:06,230
If no, has he fins?

89
00:06:07,060 --> 00:06:08,850
If he can fly that sandhog?

90
00:06:08,920 --> 00:06:11,600
If he can fly, that's a penguin if he has fins.

91
00:06:11,620 --> 00:06:12,480
Well, that's a dolphin.

92
00:06:12,490 --> 00:06:13,770
If not, that's a beer.

93
00:06:14,740 --> 00:06:17,410
Once again, this is a really basic decision.

94
00:06:17,410 --> 00:06:19,720
Trees are almost everywhere.

95
00:06:19,720 --> 00:06:23,830
So basically, when you guys go, for example, to buy food, let's say you guys are going to buy a

96
00:06:23,830 --> 00:06:24,330
pizza.

97
00:06:24,940 --> 00:06:30,340
So basically you will find decision trees when you will buy your pizza for the purposes of buying a

98
00:06:30,340 --> 00:06:30,750
pizza.

99
00:06:30,760 --> 00:06:31,990
Could be a decision tree.

100
00:06:32,950 --> 00:06:37,120
So as you can see, this is really, really used everywhere or no.

101
00:06:37,130 --> 00:06:40,870
So not only buying a pizza, so we do it without even thinking about it.

102
00:06:41,710 --> 00:06:48,850
And implementing this inside of a machine or inside of a system can be something that is there could

103
00:06:48,850 --> 00:06:54,610
be really, really useful in all aspects of our life.

104
00:06:54,610 --> 00:07:01,420
And, well, in general, simply having this inside of a system can be really useful and make the system

105
00:07:02,050 --> 00:07:04,170
more optimized and performant.

106
00:07:04,630 --> 00:07:04,900
Right.

107
00:07:04,930 --> 00:07:10,540
Another type of algorithm that we are going to talk about today will be the support vector machine,

108
00:07:11,230 --> 00:07:16,360
which is one really interesting, one really interesting algorithm.

109
00:07:16,870 --> 00:07:23,530
So basically, this is a supervised machine learning algorithm that can be used for classification or

110
00:07:23,530 --> 00:07:24,090
regression.

111
00:07:24,460 --> 00:07:30,220
But this this algorithm is mostly used for classification purposes.

112
00:07:31,150 --> 00:07:32,120
So what does this mean?

113
00:07:32,770 --> 00:07:33,700
It's pretty simple.

114
00:07:34,480 --> 00:07:43,450
So support machine is spread vector machine algorithm is able to generalise between two classes that

115
00:07:43,450 --> 00:07:49,480
are different if the set of labeled data is provided in the training set up.

116
00:07:49,960 --> 00:07:56,410
So basically, if we provide enough data in the training set of Yalgoo, I'm so basically the goal of

117
00:07:56,680 --> 00:08:03,850
this algorithm or the support vector machine algorithm will be to find a hydroplaned in an N dimensional

118
00:08:04,150 --> 00:08:06,550
space where and is the number of features.

119
00:08:06,580 --> 00:08:13,480
So basically we have features right here and in the end to be able to separate the two classes of data.

120
00:08:13,820 --> 00:08:20,970
So basically we have the two classes right here and there is a multitude of possibilities of hyper playing

121
00:08:20,980 --> 00:08:22,150
that can be chosen.

122
00:08:23,020 --> 00:08:29,260
Basically, it sounds a little bit complicated and you guys maybe don't see where exactly can be applied

123
00:08:29,260 --> 00:08:30,720
in real life.

124
00:08:31,120 --> 00:08:32,770
So let me just give you two examples.

125
00:08:32,920 --> 00:08:37,480
Let's say, for example, you want to create a facial expression classification model.

126
00:08:38,890 --> 00:08:42,790
So let's say this model will say the person is said happy.

127
00:08:43,330 --> 00:08:43,870
I don't know.

128
00:08:43,870 --> 00:08:47,750
The person is surprised, the person is shocked or whatever.

129
00:08:48,130 --> 00:08:52,930
Well, basically, you will use a support vector machine algorithm to be able to build this.

130
00:08:52,930 --> 00:09:00,460
You can do this with this type of algorithm or another thing that will be, for example, text classification.

131
00:09:00,470 --> 00:09:07,480
So basically you will compare a text written by a human being in a text written by a machine or simply

132
00:09:07,480 --> 00:09:13,230
being able to translate a text written by someone to a computer.

133
00:09:13,240 --> 00:09:19,630
So basically to work, for example, a good way to do that will be by using the support vector machine

134
00:09:20,140 --> 00:09:27,640
that will be able, with training, of course, to will be able to spot this type of text and simply

135
00:09:27,640 --> 00:09:28,180
transform it.

136
00:09:28,180 --> 00:09:36,100
So basically, there are plenty of tasks that the SVM algorithm will the SVM model or the SVM algorithm

137
00:09:36,100 --> 00:09:41,390
can realize, and it can be really useful for machine learning purposes.

138
00:09:41,650 --> 00:09:42,000
All right.

139
00:09:42,010 --> 00:09:46,330
Another algorithm that we are going to talk about today will be the night base.

140
00:09:47,200 --> 00:09:52,320
So basically this is another algorithm that is really, really used in machine learning.

141
00:09:53,440 --> 00:09:55,340
So what exactly is Navales?

142
00:09:55,780 --> 00:09:59,340
So this is a classification technique that is based on base theorem.

143
00:09:59,350 --> 00:09:59,860
So basically.

144
00:10:00,020 --> 00:10:06,080
To be able to work with it, you need to understand the basic theorem so it will assume that there is

145
00:10:06,080 --> 00:10:09,490
an independence among predictions predictors.

146
00:10:10,100 --> 00:10:17,300
So they always assume that the presence of a particular feature in a class is not related to the presence

147
00:10:17,300 --> 00:10:18,830
of any other feature.

148
00:10:19,400 --> 00:10:23,330
So this is basically the main reason why it's called naïf.

149
00:10:24,170 --> 00:10:31,130
It's it will take some features and it will use those features to be able to say, for example, OK,

150
00:10:31,130 --> 00:10:33,440
this is, I don't know, a PDA, for example.

151
00:10:34,310 --> 00:10:36,190
So let me just give you a quick example.

152
00:10:36,200 --> 00:10:43,380
So let's say something is around with 500 grams and has the different colors on it.

153
00:10:43,400 --> 00:10:49,350
So basically the theorem will say, OK, well, this algorithm will say, OK, this is a piece.

154
00:10:49,400 --> 00:10:55,850
So basically all the attributes of that, I said have the attributes of the pizza, for example.

155
00:10:56,670 --> 00:11:00,560
And so even if it's not necessarily making sense, so it could be, for example, a cake.

156
00:11:00,920 --> 00:11:05,550
It could be, for example, I don't know any other thing that is around with some colors on it.

157
00:11:06,620 --> 00:11:11,050
So basically, this is why this is the main reason why it's called naive.

158
00:11:11,240 --> 00:11:18,860
But once again, it could be really useful in classification and as well as to understand some other

159
00:11:18,860 --> 00:11:19,140
things.

160
00:11:19,160 --> 00:11:24,980
So basically, it's a well, it's a one of machine learning models.

161
00:11:25,310 --> 00:11:28,740
And basically, if it's here, it's because it does a great job.

162
00:11:29,300 --> 00:11:32,680
So basically, how is how many of these words?

163
00:11:32,710 --> 00:11:37,720
So just to give you a small introduction to how it works, you not necessarily need to understand that

164
00:11:37,730 --> 00:11:38,430
100 percent.

165
00:11:38,900 --> 00:11:45,860
So basically, you will start with calculating the calculation of all prior probabilities for a given

166
00:11:45,860 --> 00:11:52,540
class label, and then you calculate all conditional probabilities with each attribute for each class.

167
00:11:53,480 --> 00:11:57,860
Then you will multiply some class, some classes, conditional probabilities.

168
00:11:58,310 --> 00:12:03,710
You multiply prior probabilities with the number, the step three probabilities.

169
00:12:04,220 --> 00:12:11,330
And finally, you will understand what class belongs to the given input that sets them, which in other

170
00:12:11,330 --> 00:12:15,800
words, is the class that is with the higher probabilities.

171
00:12:16,790 --> 00:12:19,580
So you can have an example right here.

172
00:12:21,600 --> 00:12:27,840
All right, the next step of algorithm that we are going to talk about in this class will be the and

173
00:12:27,840 --> 00:12:29,880
then or the key nearest neighbor.

174
00:12:30,180 --> 00:12:33,630
So we say this is pretty simple to understand.

175
00:12:34,350 --> 00:12:37,390
So what exactly is this machine learning algorithm?

176
00:12:37,890 --> 00:12:42,590
So the key interesting algorithm is a type of supervised machine learning algorithm.

177
00:12:43,110 --> 00:12:49,560
So basically it can be used for regression, predictive problems, as well as for classification problems,

178
00:12:50,010 --> 00:12:54,560
but it's mainly used for classification, predictive problems.

179
00:12:54,560 --> 00:12:57,950
So basically it's mainly used for this, but it can be used for both of them.

180
00:12:57,960 --> 00:13:02,890
So it could be used for classification as well as regression problems, but it's mainly used in classification.

181
00:13:03,900 --> 00:13:10,590
So the Kinect algorithm has no specialized training phases and is using all the data for training while

182
00:13:10,590 --> 00:13:11,440
classification.

183
00:13:11,970 --> 00:13:15,780
So this is why it's really known as a lazy algorithm.

184
00:13:15,780 --> 00:13:17,760
So basically lazy learning algorithm.

185
00:13:17,770 --> 00:13:18,090
Sorry.

186
00:13:19,150 --> 00:13:26,740
So this is one of the main reasons of why the and then doesn't assume anything about underlying data,

187
00:13:26,740 --> 00:13:31,850
that is that is why it's also considered a non parametric learning algorithm.

188
00:13:31,870 --> 00:13:39,790
So basically, if there are there is two things that we can well say about the CNN algorithm is that,

189
00:13:39,970 --> 00:13:45,180
first of all, it's elizee learning algorithm.

190
00:13:45,430 --> 00:13:48,790
And finally, it's a non parametric learning algorithm.

191
00:13:49,630 --> 00:13:53,150
So it can be used in many, many places in real life.

192
00:13:53,170 --> 00:13:54,530
So this is one example.

193
00:13:54,550 --> 00:13:55,900
So basically, it's loans.

194
00:13:56,650 --> 00:13:58,400
Well, we have variables.

195
00:13:58,420 --> 00:13:59,380
We have the variable long.

196
00:13:59,380 --> 00:14:04,120
We have the age here at variable, not default and defaults, which basically the person have paid for

197
00:14:04,120 --> 00:14:05,480
it, yes or no.

198
00:14:05,860 --> 00:14:12,130
And this kid will use the distance between this point for any other point and the neighbors of this

199
00:14:12,130 --> 00:14:12,400
point.

200
00:14:12,430 --> 00:14:16,540
So basically, the neighbors of this point, the closest neighbors will be this one, this one, this

201
00:14:16,540 --> 00:14:17,560
point in this point.

202
00:14:17,800 --> 00:14:24,430
And we are able to come to certain conclusions with this algorithm at the end.

203
00:14:26,230 --> 00:14:26,550
All right.

204
00:14:26,560 --> 00:14:33,190
Another type of algorithm that we are going to talk about today will be the key means algorithm, which

205
00:14:33,190 --> 00:14:40,450
is a well known unsupervised learning algorithm that is used with machine learning.

206
00:14:41,350 --> 00:14:44,560
So basically, what exactly is the key means algorithm?

207
00:14:44,980 --> 00:14:45,790
So it's very simple.

208
00:14:45,800 --> 00:14:50,720
As I said, it's an unsupervised learning algorithm that is used to solve clustering problems.

209
00:14:51,340 --> 00:15:00,220
So basically, this Ugwu is used well, uses a very simple yet powerful procedure to classify data to

210
00:15:00,220 --> 00:15:03,120
a certain number of clusters called K.

211
00:15:03,130 --> 00:15:05,180
So basically K is a cluster.

212
00:15:05,680 --> 00:15:13,570
So the way it works at first we find the number of clusters and we assume where is the center of these

213
00:15:13,570 --> 00:15:14,140
clusters?

214
00:15:14,560 --> 00:15:17,780
So basically we don't know where exactly is the center.

215
00:15:17,800 --> 00:15:20,010
So we make an assumption at first.

216
00:15:20,040 --> 00:15:23,650
We see that's the first thing we are doing when we have identified them.

217
00:15:24,970 --> 00:15:29,890
Then we can take any object or the first key object as initial center.

218
00:15:29,910 --> 00:15:35,480
So basically to be able to find the center, well, not to find, but to assume a certain center.

219
00:15:35,500 --> 00:15:36,600
This is how we will do it.

220
00:15:37,450 --> 00:15:41,330
And then the key means algorithm will do the first.

221
00:15:41,350 --> 00:15:47,650
Well, what it will do at first, it will determine the send the central coordinates, then it will

222
00:15:47,650 --> 00:15:54,610
find the distance of each object to the center and finally group all the objects based on, well, the

223
00:15:54,610 --> 00:15:56,880
minimum of the minimum distance or.

224
00:15:58,260 --> 00:16:03,250
So if we well, usually look something like this, as you can see here, all the objects are grouped,

225
00:16:03,420 --> 00:16:06,910
grouped together and to be able to understand a little bit more.

226
00:16:06,930 --> 00:16:08,670
Let's talk about some real life example.

227
00:16:08,700 --> 00:16:15,290
So basically, this machine learning algorithm could be used in plenty of ways.

228
00:16:15,720 --> 00:16:21,830
So some of the examples in which we can use it in real life can be, for example, document classification.

229
00:16:22,170 --> 00:16:26,650
So basically we can classify different documents in different places, for example.

230
00:16:26,670 --> 00:16:31,020
So this could be a really interesting way to use this algorithm.

231
00:16:31,710 --> 00:16:34,540
Another thing, for example, could be customer segmentation.

232
00:16:34,560 --> 00:16:39,900
So basically dividing customers into different, well, types of customers, for example, customers

233
00:16:39,900 --> 00:16:44,610
that spend a lot of customers that use, I don't know, credit cards, customers that use debit cards.

234
00:16:45,270 --> 00:16:46,580
Customers that spade a spade.

235
00:16:46,590 --> 00:16:47,400
Cash customers.

236
00:16:47,400 --> 00:16:48,540
That comes once a week.

237
00:16:48,810 --> 00:16:56,010
So really, you can really Sigmon segmented your consumers and another way, for example, insurance

238
00:16:56,010 --> 00:16:56,330
fraud.

239
00:16:56,340 --> 00:17:03,900
So basically finding anomalies into, well, insurance contracts, for example, or people who are trying

240
00:17:03,900 --> 00:17:04,680
to front entrance.

241
00:17:04,680 --> 00:17:07,200
And so basically, there are plenty of ways that you can flood insurance.

242
00:17:07,270 --> 00:17:14,760
So basically, people who FROGH insurances, it's possible to well, to to be able to make a detection

243
00:17:14,760 --> 00:17:18,510
of those frauds by using the ketamine's algorithm.

244
00:17:19,530 --> 00:17:19,830
All right.

245
00:17:19,830 --> 00:17:25,650
So the other thing that we are going to talk about in this class would be the random forest algorithm,

246
00:17:25,980 --> 00:17:30,700
which is once again another really powerful algorithm that is used in machine learning.

247
00:17:31,150 --> 00:17:36,300
So for those of you who don't remember, we talked at a little bit at the beginning of this class about

248
00:17:36,300 --> 00:17:37,320
decision trees.

249
00:17:37,620 --> 00:17:42,480
And this is exactly what it's the same county, the same concept.

250
00:17:42,510 --> 00:17:46,110
So it's a concept of decision we just made bigger.

251
00:17:46,470 --> 00:17:47,730
So let me explain.

252
00:17:47,760 --> 00:17:54,900
So basically, Random Forest is a supervised machine learning algorithm that is used for classification

253
00:17:54,900 --> 00:17:55,470
and regression.

254
00:17:55,480 --> 00:18:00,480
So it can be used for both, but it's mostly used for classification problems.

255
00:18:00,780 --> 00:18:05,460
This algorithm works with decision trees and there is trees in the algorithm.

256
00:18:05,480 --> 00:18:10,530
Well, the more there are trees in the algorithm, the more the forest will be robust.

257
00:18:10,830 --> 00:18:13,890
So in other words, here, for example, we have four trees.

258
00:18:15,320 --> 00:18:19,700
So it's more powerful than if we have tree trees, but not more powerful if we have five.

259
00:18:20,400 --> 00:18:25,500
I exactly where it's basically you have decisions that are made in each tree.

260
00:18:25,530 --> 00:18:30,760
So basically decisions are made in each tree and it has to have the same outcome.

261
00:18:30,780 --> 00:18:34,420
So basically each tree needs to have a certain outcome.

262
00:18:34,440 --> 00:18:41,490
So basically choosing between A, B or C and basically each tree will choose based on different well,

263
00:18:41,550 --> 00:18:45,600
different mapping will choose a different outcome.

264
00:18:45,630 --> 00:18:52,140
So basically, the most trees, the more trees you have that choose a certain outcome, well, the more

265
00:18:52,140 --> 00:18:58,650
this outcome will be reduced or strong and the more there are chances that this outcome is the best

266
00:18:58,650 --> 00:19:06,810
outcome that you can have based on the well, based on the random forest indifference that if we only

267
00:19:06,810 --> 00:19:12,630
choose only work with one decision tree, we only have what we only will know.

268
00:19:12,660 --> 00:19:16,650
What is the be the best decision based on one decision tree?

269
00:19:17,070 --> 00:19:19,840
But with a random forest, we are able to know how.

270
00:19:19,950 --> 00:19:26,070
What is the best decision based on many trees that will each of them, each of them will map different

271
00:19:26,070 --> 00:19:27,870
scenarios in different situations.

272
00:19:28,740 --> 00:19:36,890
And at the end this will be a we better well, we better wait to find out the best answer to a problem.

273
00:19:36,900 --> 00:19:41,940
So instead of using only one decision tree using a complete random forest.

274
00:19:43,050 --> 00:19:45,980
So if we want to think about it in a certain way.

275
00:19:45,990 --> 00:19:47,200
So basically, this is just.

276
00:19:48,530 --> 00:19:53,510
A decision tree that is upgraded so we can see it this way, so basically it's more than one decision

277
00:19:53,990 --> 00:19:55,210
and it's more than one decision.

278
00:19:55,220 --> 00:19:58,670
So this will simply confirm our answer that we have in one.

279
00:19:59,600 --> 00:19:59,920
All right.

280
00:19:59,930 --> 00:20:05,990
So I hope you guys right now understood all the different types of algorithms that exist in machine

281
00:20:05,990 --> 00:20:06,340
learning.

282
00:20:06,350 --> 00:20:11,240
So those are just the examples of the most used and even the most simple.

283
00:20:12,530 --> 00:20:16,040
So once again, I'm not expecting you guys to become a professional.

284
00:20:16,040 --> 00:20:18,290
So this is just a brief introduction.

285
00:20:18,290 --> 00:20:25,040
I can spend hours talking just about each of those algorithms with all the mathematical and statistical

286
00:20:25,040 --> 00:20:26,260
aspect of each of them.

287
00:20:26,570 --> 00:20:29,040
So I try to be as straight to the point as possible.

288
00:20:29,090 --> 00:20:35,990
So right now, guys, you have a complete idea of different algorithms that exist in machine learning.

289
00:20:36,470 --> 00:20:40,280
So class goes into our next class.