1
00:00:01,750 --> 00:00:06,040
Very often we find that some values in our data are missing.

2
00:00:07,470 --> 00:00:13,710
Now, this may happen when you are collecting data from people asking them to fill a physical form and

3
00:00:13,710 --> 00:00:15,390
they do not answer all questions.

4
00:00:16,680 --> 00:00:19,540
This may happen because some of the data stored got corrupted.

5
00:00:20,850 --> 00:00:22,290
There may be many other reasons.

6
00:00:23,970 --> 00:00:30,810
But this results into a serious problem as we cannot run a machine learning algorithm on missing values.

7
00:00:32,590 --> 00:00:39,040
Now, this leaves us with only two options, either we remove the entire rule, which has some missing

8
00:00:39,040 --> 00:00:45,160
value, when we have large dataset and only few of the observations are missing value, then we can

9
00:00:45,160 --> 00:00:47,770
delete those few observations and still have.

10
00:00:48,780 --> 00:00:55,560
Substantial number of observations to print and test that model, but if the number of missing values

11
00:00:55,680 --> 00:00:59,790
is high, then the moving observations is not advisable.

12
00:01:01,200 --> 00:01:08,790
In such a case, we should replace the blank places with some harmless values, harmless meaning that

13
00:01:08,790 --> 00:01:10,740
they should not impact the model.

14
00:01:10,740 --> 00:01:19,770
Much such neutral values are values close to the center of that variable so we can choose mean or median

15
00:01:19,770 --> 00:01:21,990
value to people in these blanks.

16
00:01:24,070 --> 00:01:28,000
Here are some most common ways to do missing value imputation.

17
00:01:29,860 --> 00:01:36,070
First is to impute with zero note that a missing value is not zero.

18
00:01:36,310 --> 00:01:39,400
You can simply put a zero where you have a missing value.

19
00:01:40,890 --> 00:01:46,830
But this really makes business sense, so if zero makes sense in any particular scenario, we can do

20
00:01:46,830 --> 00:01:47,100
that.

21
00:01:48,960 --> 00:01:53,660
Second option is to imbued with centers as this will not have much impact on the model.

22
00:01:54,990 --> 00:02:00,690
For example, if we have high rates of people as a variable and some of the values are missing in the

23
00:02:00,690 --> 00:02:03,990
data, putting zero does not make any sense.

24
00:02:05,050 --> 00:02:09,160
So we can put mean or median height of people in the missing field.

25
00:02:10,990 --> 00:02:17,260
If you have a categorical variable with some values missing, you can assign that category which has

26
00:02:17,260 --> 00:02:19,890
maximum frequency in the missing field.

27
00:02:22,420 --> 00:02:28,860
So suppose we have gender data and 90 percent of them are males, then in the field of the variable

28
00:02:29,020 --> 00:02:30,820
we can safely put male.

29
00:02:32,950 --> 00:02:39,340
The third option is, again, putting in mean value, but instead of using the population mean, if

30
00:02:39,340 --> 00:02:45,160
we can make out relevant segments from the data we can use this segment mean for that segments observation.

31
00:02:46,280 --> 00:02:52,820
For example, if we have rainfall data of different cities, but the data is missing for some cities

32
00:02:53,060 --> 00:03:00,950
instead of using population mean we can take the mean of neighboring cities or cities belonging to similar

33
00:03:00,950 --> 00:03:05,330
regions or cities belonging to same state to fill the blanks.

34
00:03:06,540 --> 00:03:10,580
This makes sense since rainfall does not worry a lot among neighboring cities.

35
00:03:12,700 --> 00:03:19,960
Therefore, we should use business knowledge so that we can select one of these methods wisely.

36
00:03:21,820 --> 00:03:29,930
When we find it in any software, the softwares do give us the number of blank values in each variables.

37
00:03:31,180 --> 00:03:35,090
We will learn how to fill the blank values in this software.

38
00:03:35,110 --> 00:03:36,040
We are going to use.