1
00:00:00,990 --> 00:00:06,930
Now, once our data is ready, the first step that we take is to split the data into two parts.

2
00:00:08,250 --> 00:00:10,800
The first part will be called the training set.

3
00:00:11,790 --> 00:00:18,440
This part will be used to train the model and then that will be a second set.

4
00:00:18,930 --> 00:00:26,160
That second set will be Kordic tested and that test set will be used to test the performance of the

5
00:00:26,160 --> 00:00:26,940
created model.

6
00:00:28,920 --> 00:00:36,720
So in business and adio, if you have a lot of data, it is always recommended that you train the model

7
00:00:36,990 --> 00:00:42,900
on one set of observations and you keep one set of observations which are not shown into your model.

8
00:00:43,710 --> 00:00:48,840
You test your performance or your offer model on this previously unseen data.

9
00:00:50,880 --> 00:00:54,380
Usually we do a split of 80 20.

10
00:00:54,870 --> 00:01:01,530
That is, we use 80 percent of the data available to train the model and 20 percent of the remaining

11
00:01:01,530 --> 00:01:04,350
data to test the performance.

12
00:01:05,250 --> 00:01:13,710
So here I am showing you how to split the available data into two parts to split the data in are we

13
00:01:13,710 --> 00:01:16,380
use this package called See It tools.

14
00:01:19,470 --> 00:01:28,020
If you have this package in stolen art, it will be visible here in the packages part so you can scroll

15
00:01:28,050 --> 00:01:30,960
and see if see it is available with you or not.

16
00:01:31,230 --> 00:01:33,370
If it is not your undies, come on in.

17
00:01:33,450 --> 00:01:34,770
Stalled out packages.

18
00:01:36,250 --> 00:01:37,250
See Édouard.

19
00:01:38,520 --> 00:01:42,210
No doubt see it tools is written in single quotation marks.

20
00:01:43,460 --> 00:01:44,420
So we'll run this command.

21
00:01:49,730 --> 00:01:53,090
And it has installed seer tools in the packages.

22
00:01:53,270 --> 00:01:55,520
You can see that see a tool is available here.

23
00:01:57,080 --> 00:02:00,530
Now it is installed, but it is not active.

24
00:02:00,680 --> 00:02:02,000
That is, it cannot be used.

25
00:02:02,060 --> 00:02:03,860
As of now to make it active.

26
00:02:03,890 --> 00:02:05,330
We run this library command.

27
00:02:06,130 --> 00:02:10,580
You can also take this checkbox to will run this library command.

28
00:02:12,650 --> 00:02:14,510
And you can see that a tick comes here.

29
00:02:16,490 --> 00:02:19,520
And this year, tools package is now active.

30
00:02:23,920 --> 00:02:26,210
The next step is setting seed.

31
00:02:27,960 --> 00:02:32,280
We set seed so that we have reproducible results.

32
00:02:34,000 --> 00:02:42,370
What I mean by this is when we are doing a test transplant, we'll be randomly putting 80 percent of

33
00:02:42,370 --> 00:02:46,450
the data into printed and 20 percent of the data to test it.

34
00:02:47,800 --> 00:02:55,390
If we do not succeed, my randomly selected 80 percent of the data will be different than your randomly

35
00:02:55,480 --> 00:02:57,190
selected 80 percent of the data.

36
00:02:58,510 --> 00:03:01,090
But if we set seed to zero.

37
00:03:01,810 --> 00:03:07,990
If you also said the seed to zero and if I have said the seed to zero here, both of us will be getting

38
00:03:07,990 --> 00:03:09,310
these same split of data.

39
00:03:12,140 --> 00:03:19,670
And if we both get same speed of test and train data, we both will get the same model and that same

40
00:03:19,670 --> 00:03:21,740
model will be giving us these same results.

41
00:03:24,310 --> 00:03:27,460
So that is why we are setting the seed to zero hit.

42
00:03:31,140 --> 00:03:35,880
Next will create a variable called split in the split variable.

43
00:03:36,390 --> 00:03:37,710
We will have value.

44
00:03:37,780 --> 00:03:38,880
True and false.

45
00:03:39,690 --> 00:03:41,360
And we are giving a split ratio.

46
00:03:41,430 --> 00:03:48,870
Point eight, which means that 80 percent of the values will be true and 20 percent will be false.

47
00:03:50,730 --> 00:03:57,990
So if I run this command, you can see that a split variable is created here, which contains value

48
00:03:58,220 --> 00:04:00,450
through brute rule force and so on.

49
00:04:01,470 --> 00:04:08,550
And randomly, nearly 80 percent of these values will be true and nearly 20 percent will be false.

50
00:04:10,260 --> 00:04:17,520
Now, in my training said, I'll be taking all the movie data set values which have split value through.

51
00:04:19,350 --> 00:04:24,710
And indeed, they said, I'll have all the movie data set values which have split value falls.

52
00:04:26,190 --> 00:04:32,190
I've named this variable as Trinity because for now we are doing classification.

53
00:04:34,230 --> 00:04:39,020
That is, we are trying to predict a very well with this categorical later on in the course.

54
00:04:39,180 --> 00:04:41,740
We'll also see how to do regression in swim.

55
00:04:43,200 --> 00:04:47,370
So here I have created two variables Rainsy and D.C. Green.

56
00:04:47,440 --> 00:04:52,390
He will get values from what we dataset, which has split value through.

57
00:04:52,630 --> 00:04:53,580
So at underscore my.

58
00:04:55,320 --> 00:05:01,590
And you can see nearly 80 percent, which is nearly 400 observations are getting stored in the trains

59
00:05:01,660 --> 00:05:02,120
very well.

60
00:05:02,940 --> 00:05:05,850
And I also found this one, which is DWC.

61
00:05:06,740 --> 00:05:12,060
And I have one hundred eight observations in my test dataset.

62
00:05:13,100 --> 00:05:21,040
This is how we split the data in two days and Drame will be using the brain to do does it to rain as

63
00:05:21,130 --> 00:05:21,690
a model.

64
00:05:22,100 --> 00:05:27,550
And later on, we will compare its performance on the test see dataset.
