1
00:00:01,870 --> 00:00:06,320
In this session, we will understand how to create a decision tree.

2
00:00:07,310 --> 00:00:16,790
Decision three forms the basis for random for the next most two most widely used algorithms in the world

3
00:00:16,790 --> 00:00:17,110
today.

4
00:00:18,430 --> 00:00:20,910
So let's start this with an example.

5
00:00:22,080 --> 00:00:28,260
In this example, I am trying to predict the likelihood of an individual getting a heart disease and

6
00:00:28,260 --> 00:00:33,620
I am considering three factors for that blood pressure, sugar and cholesterol.

7
00:00:34,350 --> 00:00:38,550
So heart disease, the likelihood of someone getting a heart disease.

8
00:00:39,570 --> 00:00:45,090
It's my dependent variable, and these three are independent variables.

9
00:00:46,000 --> 00:00:46,470
OK?

10
00:00:47,570 --> 00:00:51,210
Now, a decision tree is constructed like this.

11
00:00:51,390 --> 00:00:58,080
We start with Amy, one of the three factors, and then go on constructing a tree like this high sugar.

12
00:00:58,430 --> 00:00:59,960
Is it true or false?

13
00:01:00,620 --> 00:01:03,440
If it is true means that the individual IBP.

14
00:01:04,220 --> 00:01:04,580
Yes.

15
00:01:04,580 --> 00:01:09,530
The individual at BP, did he also have high cholesterol, so on and so forth.

16
00:01:10,620 --> 00:01:13,650
The challenge in Decision Tree is determining.

17
00:01:15,070 --> 00:01:20,020
Where to start from, that is, which factors should I start the decision tree from?

18
00:01:21,170 --> 00:01:28,490
That is known as root, not root, nor is the starting point in a decision tree data mining route known

19
00:01:28,490 --> 00:01:28,850
as the.

20
00:01:29,820 --> 00:01:36,360
First steps in constructing the decision, because you can start with a sugar high or high cholesterol,

21
00:01:36,370 --> 00:01:36,670
right?

22
00:01:36,870 --> 00:01:38,990
So I must know where to start from.

23
00:01:39,480 --> 00:01:46,560
That is done using two is one is Guinea index and the other is known as Entropia information.

24
00:01:47,460 --> 00:01:47,910
OK.

25
00:01:49,240 --> 00:01:58,120
Now let's see how to compute the group not using Guiney index, so I have populated some values based

26
00:01:58,120 --> 00:01:59,330
on historical data.

27
00:01:59,980 --> 00:02:05,510
Someone I am considering, if you see, I am considering each of the three factors in isolation now.

28
00:02:05,810 --> 00:02:12,250
OK, so I will consider the factors in isolation for the purpose of determining which node to start

29
00:02:12,250 --> 00:02:12,520
from.

30
00:02:12,950 --> 00:02:22,720
Right so high BP when I BP was through the likelihood of someone developing heart disease out of so

31
00:02:22,720 --> 00:02:30,070
many people who had I be one of five people, developed heart disease and 39 did not develop our disease.

32
00:02:31,840 --> 00:02:41,680
Out of so many people not having I be 34 people developed heart disease and 125 people did not develop

33
00:02:41,680 --> 00:02:42,400
heart disease.

34
00:02:45,040 --> 00:02:51,320
Same is the case with high sugar and high cholesterol, so many people had high cholesterol that refused

35
00:02:51,320 --> 00:02:55,060
to prove right out of their mind.

36
00:02:55,100 --> 00:03:00,110
Two people develop heart disease and 31 did not develop heart disease.

37
00:03:01,190 --> 00:03:03,660
And some people did not have high cholesterol.

38
00:03:03,830 --> 00:03:11,330
That is why you have Olsher out of them fortified developed heart disease and 129 did not develop heart

39
00:03:11,330 --> 00:03:11,740
disease.

40
00:03:12,590 --> 00:03:12,870
Right.

41
00:03:13,190 --> 00:03:17,410
So this representation is what you do first, right?

42
00:03:18,510 --> 00:03:24,900
And then you compute Guiney index using the Formula One minus probability of Estherville square one

43
00:03:24,900 --> 00:03:27,790
minus minus probability of false score.

44
00:03:28,380 --> 00:03:32,470
OK, so I, I'm explaining this for BP, right.

45
00:03:32,730 --> 00:03:35,840
I now find the son will not find us.

46
00:03:36,030 --> 00:03:40,650
Thirty nine total is one forty four and similarly total is 159.

47
00:03:40,920 --> 00:03:44,700
OK, so I will compute Guinea for two.

48
00:03:44,700 --> 00:03:45,870
Guinea for false.

49
00:03:46,890 --> 00:03:47,260
Right.

50
00:03:47,790 --> 00:03:56,290
So how will I do four to one minus one, not five, divided by thirty nine, plus one on five.

51
00:03:56,460 --> 00:03:59,310
That is one four for the whole square minus no scenario.

52
00:04:00,180 --> 00:04:02,270
I get a value of point three nine five.

53
00:04:02,760 --> 00:04:08,460
And then for this I get a value off point three three six fast I consider it.

54
00:04:08,460 --> 00:04:08,970
Yes.

55
00:04:09,180 --> 00:04:16,590
Thirty four divided by one fifty nine which is thirty four plus one twenty five then I consider no which

56
00:04:16,590 --> 00:04:19,830
is one twenty five thirty four plus one twenty five.

57
00:04:19,830 --> 00:04:21,060
Is that in the denominator.

58
00:04:21,070 --> 00:04:22,950
And then I square it right.

59
00:04:23,100 --> 00:04:25,070
I got getting for true and false.

60
00:04:25,080 --> 00:04:28,050
Now let's find the total need for BP.

61
00:04:28,440 --> 00:04:35,370
That is one forty four for the two and one fifty nine for the Fox.

62
00:04:36,030 --> 00:04:36,360
Right.

63
00:04:36,630 --> 00:04:37,500
I multiply.

64
00:04:39,130 --> 00:04:43,480
This value with the three nine five one three three six that I want.

65
00:04:44,470 --> 00:04:48,190
If I then have total meaning for BP.

66
00:04:49,440 --> 00:04:50,110
Is this clear?

67
00:04:52,060 --> 00:04:52,500
OK.

68
00:04:55,440 --> 00:05:03,480
I will use the same approach to compute Guinea for sugar and cholesterol, also frosti compute Guinea

69
00:05:03,480 --> 00:05:08,640
for to Guinea for false and then I compute totally.

70
00:05:09,000 --> 00:05:09,330
Right.

71
00:05:09,810 --> 00:05:14,820
So I have taught in Guinea for BP sugar and cholesterol.

72
00:05:15,360 --> 00:05:23,820
OK, in the case of BP, this point, three six four sugar is point three six point three eight one.

73
00:05:24,690 --> 00:05:31,710
So that will start the decision from I will start where the guinea is the lowest.

74
00:05:32,550 --> 00:05:36,890
OK, Guinea is lowest in the case of sugar.

75
00:05:36,900 --> 00:05:40,370
So my will start with sugar.

76
00:05:40,830 --> 00:05:43,110
The next will be the next highest one.

77
00:05:43,440 --> 00:05:43,790
Right.

78
00:05:44,040 --> 00:05:46,620
That is BP and then cholesterol.

79
00:05:46,620 --> 00:05:48,560
Where are you getting it.

80
00:05:48,870 --> 00:05:52,100
So this is how you construct the decision tree.

81
00:05:52,950 --> 00:06:00,120
If you are wondering, do you need to do so many computations, so many calculations for creating a

82
00:06:00,120 --> 00:06:01,440
decision tree algorithm?

83
00:06:02,400 --> 00:06:10,080
Don't worry, there are pre-built libraries available in Python that makes all these computations very

84
00:06:10,080 --> 00:06:10,470
easy.

85
00:06:10,680 --> 00:06:15,660
OK, you just need to identify the dependent and independent variables.

86
00:06:16,290 --> 00:06:24,330
The process of developing the tree, fitting the model is done by the library that makes the job so

87
00:06:24,330 --> 00:06:24,870
much easier.

88
00:06:24,870 --> 00:06:25,190
Right?

89
00:06:26,190 --> 00:06:30,800
OK, we used to Guinea for the purpose of identifying the root.

90
00:06:31,620 --> 00:06:33,720
We can also use entropy.

91
00:06:34,320 --> 00:06:40,740
OK, mostly the root node will be the same both in Guinea as well as an entropy.

92
00:06:41,070 --> 00:06:43,920
In few cases they will differ.

93
00:06:44,920 --> 00:06:52,160
Right, the route will differ between Guinea and entropy in a few cases.

94
00:06:53,180 --> 00:06:53,600
OK.