1
00:00:01,990 --> 00:00:06,100
Let's see this important concept, outliers.

2
00:00:07,700 --> 00:00:18,620
Outliers are basically extreme points in our data set, and they impact the accuracy of our models quite

3
00:00:18,620 --> 00:00:19,190
adversely.

4
00:00:20,240 --> 00:00:29,240
This graph you will remember from the central tendency and dispersion session, I use this to explain

5
00:00:29,240 --> 00:00:31,020
the concept of standard deviation.

6
00:00:31,590 --> 00:00:35,510
I'm now going to use the same graph to explain outliers.

7
00:00:36,230 --> 00:00:38,630
So what are outliers in this example?

8
00:00:39,910 --> 00:00:47,650
If you really see the outliers at this point, sixty nine, seventy one and 96.

9
00:00:49,040 --> 00:00:53,510
So the question is, can we have outliers in our dataset?

10
00:00:54,550 --> 00:00:55,420
You can have.

11
00:00:56,290 --> 00:01:00,340
Provided the number of such outliers are negligible.

12
00:01:02,570 --> 00:01:09,150
And a better way to handle outliers is to remove them, remove outliers from your dataset.

13
00:01:09,680 --> 00:01:11,300
Why do you want to have outliers?

14
00:01:12,050 --> 00:01:16,130
Outliers can lead to low accuracy.

15
00:01:17,860 --> 00:01:24,100
Unwanted inferences can result if outliers are present in our dataset.

16
00:01:24,700 --> 00:01:31,150
OK, we're going to use more explored to understand outliers.

17
00:01:31,720 --> 00:01:33,760
Your box plot will look like this.

18
00:01:34,210 --> 00:01:38,050
The outliers will shown as a daughter star mark.

19
00:01:38,320 --> 00:01:42,550
OK, and the limits for box plot are.

20
00:01:44,530 --> 00:01:51,550
Q three plus one point five times you three minus one, are you one minus one point five times three

21
00:01:51,550 --> 00:01:59,790
minus Q on this one point five times three minus one is also known as important range.

22
00:02:00,670 --> 00:02:05,560
OK, unwater Q1 and Q3, they are percentiles.

23
00:02:06,310 --> 00:02:10,600
I'm sure you will remember percent from the previous session.

24
00:02:11,440 --> 00:02:17,980
I'm going to use Excel to demonstrate the concept of outliers and box plus.

25
00:02:19,910 --> 00:02:25,760
You can use Python or any other programming language to do the same.

26
00:02:26,390 --> 00:02:28,760
OK, now let's see.

27
00:02:29,810 --> 00:02:31,150
Box plot in action.

28
00:02:31,160 --> 00:02:34,330
I have the invoice amount for this.

29
00:02:34,340 --> 00:02:36,740
I'm going to create a box plot.

30
00:02:36,830 --> 00:02:39,050
OK, how many data points are there?

31
00:02:39,530 --> 00:02:46,490
There are 156 data points in our lives to let them come to insert.

32
00:02:47,760 --> 00:02:49,410
Click recommended charts.

33
00:02:51,100 --> 00:02:58,840
Come to all charts, click Moxham Whiskered, OK, so this is your box plot.

34
00:03:00,340 --> 00:03:05,880
So as you can see here, the outliers are clearly shown as dots, right?

35
00:03:06,760 --> 00:03:08,980
It is also telling what are those values?

36
00:03:10,720 --> 00:03:11,670
So we can remove them.

37
00:03:12,710 --> 00:03:15,200
Removed outliers from the box plot.

38
00:03:16,550 --> 00:03:20,210
OK, and as I mentioned earlier, you can use.

39
00:03:21,360 --> 00:03:28,480
Python or any of the programming language to create a box plot and to remove the outliers also.

40
00:03:29,290 --> 00:03:31,500
OK, you can do that programmatically.

41
00:03:33,150 --> 00:03:33,520
Right.

42
00:03:33,930 --> 00:03:41,440
So, again, a very powerful concept, as I said, outliers come in the way of our accuracy.

43
00:03:41,460 --> 00:03:44,820
So let's remove them to the extent possible.

44
00:03:45,870 --> 00:03:46,290
OK.
