1
00:00:02,270 --> 00:00:07,850
When we run the code, we need to specify the stopping criteria for arbitrary.

2
00:00:09,620 --> 00:00:12,530
By default, programs set some criteria.

3
00:00:13,160 --> 00:00:16,790
But we need to understand this so that we can control the dryland.

4
00:00:18,540 --> 00:00:22,430
Also, one of the major problems with Rhys's overfitting.

5
00:00:24,030 --> 00:00:28,440
This problem can be avoided if we do not let the tree grow beyond a certain point.

6
00:00:31,410 --> 00:00:36,420
So here I am showing you three methods which we can use to control tree growth.

7
00:00:39,750 --> 00:00:45,960
First is by specifying the minimum number of observations required to split the node.

8
00:00:47,470 --> 00:00:52,510
Suppose we said that we need minimum six observations at a node.

9
00:00:53,520 --> 00:00:54,280
Most splitting.

10
00:00:55,540 --> 00:00:59,080
So if you remember that example, we had 10 student.

11
00:01:00,050 --> 00:01:06,650
We made a split and we got five students in this node and five students in this.

12
00:01:08,480 --> 00:01:16,060
Now, if I had a criteria that minimum number of observations in a. node should be six to consider a

13
00:01:16,060 --> 00:01:17,390
split in this node.

14
00:01:18,410 --> 00:01:25,880
Both of these nodes will not meet the criteria and the regrowth will stop after the first played only.

15
00:01:28,020 --> 00:01:34,500
The next method is specifying the minimum number of observations required at the leaf node.

16
00:01:36,500 --> 00:01:44,840
So when we consider making a split after the split, we can specify how many minimum number of observations

17
00:01:45,050 --> 00:01:48,260
should be present in each node after that split.

18
00:01:49,310 --> 00:01:55,080
So suppose if I said that after a split, there should be at least three observations.

19
00:01:56,720 --> 00:02:02,990
Then when I'm splitting this node and I'm getting three of the revision, then this and only two in

20
00:02:02,990 --> 00:02:07,040
this, this leaf node will not meet the criteria.

21
00:02:07,880 --> 00:02:10,460
Therefore, this split will not be possible.

22
00:02:10,580 --> 00:02:15,290
As leaf node is not meeting the criteria which I am setting, I did leave node.

23
00:02:16,320 --> 00:02:20,760
So this is setting the value of minimum observations in the leaf node.

24
00:02:23,330 --> 00:02:24,380
The third method is.

25
00:02:25,370 --> 00:02:28,190
Giving the maximum depth of the three.

26
00:02:29,940 --> 00:02:32,380
So ar3 started, Zeder, that.

27
00:02:32,580 --> 00:02:34,050
So this Naude is zero.

28
00:02:34,770 --> 00:02:42,420
If I mentioned that I need a tree with a maximum depth of one, it will do one split, which is splitting

29
00:02:42,420 --> 00:02:43,730
on the basis of ours.

30
00:02:44,220 --> 00:02:45,720
And after that, it will stop.

31
00:02:47,390 --> 00:02:51,920
If I mentioned to this is depth zero, this is depth one.

32
00:02:52,610 --> 00:02:53,960
It will again do a split.

33
00:02:54,470 --> 00:02:55,740
This will be depth two.

34
00:02:56,030 --> 00:02:57,740
And after that, it will stop.

35
00:02:58,900 --> 00:03:03,070
So using maximum depth, you get to control the growth of this tree.

36
00:03:04,390 --> 00:03:12,480
So using these three method before you run your decision tree, you can tell that at which step rebuilding

37
00:03:12,520 --> 00:03:14,010
process should stop.