1
00:00:00,770 --> 00:00:09,660
A simple and a popular method for removing untrained and a seasonal pattern from a cities is by the

2
00:00:09,660 --> 00:00:11,360
operation of different thing.

3
00:00:13,320 --> 00:00:16,650
Different thing means taking the difference between two values.

4
00:00:18,330 --> 00:00:19,270
So a lag.

5
00:00:19,290 --> 00:00:27,000
One difference, which is also called first difference, means that we are taking difference between

6
00:00:27,180 --> 00:00:30,000
every two consecutive values in the series.

7
00:00:31,840 --> 00:00:39,590
For example, in this snippet of a line mice data, you can see that in this second column we have a

8
00:00:39,590 --> 00:00:43,470
line Miles traveled in the third column.

9
00:00:43,710 --> 00:00:46,650
We have lag one values of these data points.

10
00:00:47,310 --> 00:00:52,140
So the mileage value of January is now present in the second row.

11
00:00:53,580 --> 00:00:56,670
The mileage value of fab is now present in the third row.

12
00:00:57,090 --> 00:01:01,200
So this is basically lag one value of this Miles column.

13
00:01:02,940 --> 00:01:04,320
When we difference these two.

14
00:01:05,310 --> 00:01:10,770
So if you look at the second row, the difference between these two values is minus six.

15
00:01:10,770 --> 00:01:14,130
Forty nine for the third row.

16
00:01:14,240 --> 00:01:15,930
The defense's nine zero six.

17
00:01:16,320 --> 00:01:20,430
So there's different values, creates another series.

18
00:01:21,030 --> 00:01:25,080
And this series is called Lag One, Different Cities.

19
00:01:27,360 --> 00:01:37,410
Similarly, if you want to generalize, we can say that different thing at laggy is subtracting devalue

20
00:01:37,590 --> 00:01:39,120
from gay periods back.

21
00:01:40,620 --> 00:01:42,390
So it is basically whity.

22
00:01:42,540 --> 00:01:49,380
That is pantry's value at twenty minus times, these value at Bain B minus key.

23
00:01:51,390 --> 00:01:55,200
So for example, for our daily series lags seven.

24
00:01:55,200 --> 00:01:59,190
Defensing means subtracting from each value.

25
00:01:59,910 --> 00:02:02,850
The value of same day in the previous week.

26
00:02:05,740 --> 00:02:13,330
This technique of defending is used to remove print and seasonal patterns and obtain cities which like

27
00:02:13,480 --> 00:02:14,500
both of these things.

28
00:02:15,680 --> 00:02:17,140
So let's see how we remove print.

29
00:02:20,470 --> 00:02:25,810
And if we got on the top, you can see that our times this has a clear upward trend.

30
00:02:27,520 --> 00:02:34,600
When we do a flag, one different thing and bloddy lag one difference values these values.

31
00:02:34,690 --> 00:02:37,080
Now do not have any visible print.

32
00:02:38,650 --> 00:02:47,980
And you can also notice that this method of lag when the printing does not assume that the trend is

33
00:02:47,980 --> 00:02:48,430
global.

34
00:02:49,150 --> 00:02:54,290
So even if in the initial part of this series you had a slow upward trend.

35
00:02:54,760 --> 00:02:58,350
And in the later part of the series, the upward trend increases.

36
00:02:59,230 --> 00:03:03,530
Still, your lag when the printing will give you beat rendered cities.

37
00:03:05,140 --> 00:03:12,750
When you do one time different thing, you'll be able to remove linear drinks for quadratic and exploring

38
00:03:12,770 --> 00:03:13,340
children.

39
00:03:15,010 --> 00:03:22,320
We often have another round of lag when the printing, which has to be applied to our different series.

40
00:03:23,650 --> 00:03:29,320
So for a quadratic print, we will take the new series and do different thing again.

41
00:03:30,990 --> 00:03:35,050
Does it is that we will get then will not have the quadratic ring.

42
00:03:36,830 --> 00:03:39,010
Once you have a series which does not have print.

43
00:03:39,760 --> 00:03:47,380
You can use different ink to remove seasonality also for removing a seasonal pattern with M seasons

44
00:03:48,390 --> 00:03:49,690
with different at a lag.

45
00:03:49,900 --> 00:03:59,580
And for example, to remove a monthly pattern in our monthly data, we can take lag where difference.

46
00:04:00,940 --> 00:04:09,670
This means that if your sales undergo a cycle within one year, that is, for example, in the month

47
00:04:09,670 --> 00:04:16,780
of June and July, you have increasing and nearly closing of the year your sales dip in such scenarios.

48
00:04:17,230 --> 00:04:25,750
We can say that there is a yearly cycle and seasonality is monthly to remove this kind of seasonality.

49
00:04:26,470 --> 00:04:28,090
We will do a lag.

50
00:04:28,480 --> 00:04:29,080
Defensing.

51
00:04:29,860 --> 00:04:38,830
So you will different the sales value of July 2020 with the sales value of July 2019.

52
00:04:40,870 --> 00:04:47,680
The resulting series that you will get will be be seasonal least seasonality will be removed from it.

53
00:04:48,460 --> 00:04:56,230
If you have weekly seasonality, that is, if on weekends your sales increase and on weekdays your sales

54
00:04:56,230 --> 00:04:56,770
decrease.

55
00:04:57,580 --> 00:04:58,750
Then we can do a lag.

56
00:04:58,750 --> 00:05:06,550
Seven different thing in which we will subtract the sales of last Sunday from this Sunday.

57
00:05:10,690 --> 00:05:18,820
You can see in this graph also when we apply these seasonality on this left graph, we get this new

58
00:05:18,820 --> 00:05:23,530
graph which is not showing any particular cyclic seasonal effect.

59
00:05:24,310 --> 00:05:27,600
The seasonal pattern is more prominent in the initial part of the series.

60
00:05:28,090 --> 00:05:37,210
And you can see that here the seasonal pattern is now removed in the final cities in case you have bought

61
00:05:37,360 --> 00:05:39,310
trend and seasonality in your data.

62
00:05:39,880 --> 00:05:44,080
You have to do different thing twice, once to remove the trend.

63
00:05:44,620 --> 00:05:53,020
And second, to remove these seasonality in this way from the initial cities where you have both tendencies,

64
00:05:53,020 --> 00:06:01,780
Naledi, and you cannot apply many of our methods which only apply to cities which do not have prevented

65
00:06:01,790 --> 00:06:02,260
naledi.

66
00:06:02,950 --> 00:06:08,110
We apply this method of defensing to get a series which do not have random seasonality.

67
00:06:08,410 --> 00:06:13,540
And then we can apply those models where the cities must be D seasonal later.

68
00:06:13,630 --> 00:06:14,110
Detering.