1
00:00:02,020 --> 00:00:05,700
Now we will discuss a concept called seasonality in data.

2
00:00:07,680 --> 00:00:14,820
When we have time data and our data shows recurring patterns such as high is coming every three months.

3
00:00:15,870 --> 00:00:23,570
Other low coming in a particular season, such type of data is called seasonal data can have seasonality

4
00:00:23,600 --> 00:00:24,980
because of the weather.

5
00:00:25,700 --> 00:00:26,930
For example, every year.

6
00:00:27,020 --> 00:00:30,350
Ice cream sales are higher in summer and lower in winter.

7
00:00:32,130 --> 00:00:35,100
Or seasonality for tourism industry.

8
00:00:36,470 --> 00:00:38,870
Depends on holidays and vacations.

9
00:00:40,720 --> 00:00:46,810
Now, the variance in sales due to this season of the year may not be explained well by other factors

10
00:00:46,810 --> 00:00:49,370
of model and hence our model will not fit well.

11
00:00:51,440 --> 00:00:55,460
Therefore, it is advisable that we remove seasonality from the data.

12
00:00:56,830 --> 00:01:03,760
To do that, we usually find a correction factor, which we multiply with the data to get a normalized

13
00:01:03,760 --> 00:01:04,240
value.

14
00:01:06,170 --> 00:01:08,970
To calculate this factor, we can use this formula.

15
00:01:09,710 --> 00:01:13,730
M is equal to mean of a year divided by means of month.

16
00:01:15,270 --> 00:01:18,780
This will give us multiplication factor for each observation.

17
00:01:20,270 --> 00:01:22,700
Let me show you an example of how this is done.

18
00:01:26,370 --> 00:01:34,010
Here I have data off sales in the last three years and every year in the first six months, sales show

19
00:01:34,010 --> 00:01:36,080
a rise and then a dip.

20
00:01:37,340 --> 00:01:41,240
You can look at this graph from month one two, month three.

21
00:01:41,300 --> 00:01:42,140
There's a race.

22
00:01:43,260 --> 00:01:44,460
Month three and four.

23
00:01:44,650 --> 00:01:45,410
There's a flood.

24
00:01:46,000 --> 00:01:47,930
And after month food, there is a dip.

25
00:01:51,440 --> 00:01:57,080
After understanding the business context, if we are sure that this is due to the seasonal nature of

26
00:01:57,080 --> 00:02:01,730
sales, we find the factors to normalize these sales figures.

27
00:02:05,270 --> 00:02:07,370
If we use the formula shown earlier.

28
00:02:08,700 --> 00:02:11,940
I will get multiplication factors of each value.

29
00:02:15,050 --> 00:02:16,970
So this valley of 21.

30
00:02:18,270 --> 00:02:20,010
We'll get multiplication factor.

31
00:02:21,230 --> 00:02:24,920
By dividing mean of this year, by means of this month.

32
00:02:27,810 --> 00:02:32,410
But because the mean of these three years is almost similar.

33
00:02:34,580 --> 00:02:42,230
I have used population mean and divided by month, meaning I would suggest that you try doing it for

34
00:02:42,320 --> 00:02:43,010
each cell.

35
00:02:45,080 --> 00:02:52,310
So using population means divided by month, I mean, I have found these factors here, which I have

36
00:02:52,310 --> 00:02:54,030
multiplied with these sales figures.

37
00:02:56,350 --> 00:02:59,410
And below in this table, I have these multiplied values.

38
00:03:00,940 --> 00:03:06,750
So you can see if you multiply 21 with one point five two, you get thirty one point nine four.

39
00:03:07,920 --> 00:03:12,330
If you mind about twenty three with one point five two, you'll get thirty four point nine nine.

40
00:03:13,020 --> 00:03:16,590
So this is all this table is created using the multiplication factors.

41
00:03:17,670 --> 00:03:23,920
You can see that these old values are now in the same range and data is not seasonal.

42
00:03:26,020 --> 00:03:31,390
Not this data can be used for analysis and can be fed to our model.

43
00:03:32,590 --> 00:03:34,280
So this is all to move seasonality.
