1
00:00:00,000 --> 00:00:04,837
[MUSIC]

2
00:00:04,837 --> 00:00:07,969
In this module we address
a very fundamental concept,

3
00:00:07,969 --> 00:00:09,950
a concept of having missing data.

4
00:00:09,950 --> 00:00:14,935
And missing data can impact us both in
the training time And a prediction time.

5
00:00:14,935 --> 00:00:19,635
For both cases, we explored
fundamental ideas that are useful for

6
00:00:19,635 --> 00:00:22,475
a wide range of algorithms,
not just decision trees.

7
00:00:22,475 --> 00:00:27,205
We explore the idea of just skipping
data points, which has its benefits and

8
00:00:27,205 --> 00:00:32,920
pitfalls, the idea of trying to impute or
guess what those missing values are.

9
00:00:32,920 --> 00:00:35,991
And the idea of modifying
the actual learning algorithm,

10
00:00:35,991 --> 00:00:41,060
in particular with decision trees,
in order to better deal with missing data.

11
00:00:41,060 --> 00:00:46,480
Now, in practice,
you will often see missing data and

12
00:00:46,480 --> 00:00:49,030
you should be always on the lookout for
missing data.

13
00:00:49,030 --> 00:00:55,410
And sometimes our data comes in, in a way,
value is not just explicitly missing.

14
00:00:55,410 --> 00:00:59,120
So for example, sometimes people
put in zero, when it's unknown.

15
00:00:59,120 --> 00:01:00,330
And you might think it's zero.

16
00:01:00,330 --> 00:01:01,420
But it's really unknown.

17
00:01:01,420 --> 00:01:03,980
So you should always be on the lookout for
missing data.

18
00:01:03,980 --> 00:01:06,180
And you should always
take it very carefully,

19
00:01:06,180 --> 00:01:09,670
because it really impact
the answers of your algorithm.

20
00:01:09,670 --> 00:01:12,260
Today we've seen some basic
approach dealing with that.

21
00:01:12,260 --> 00:01:15,340
Of course, there are more advanced
ones that you can get into.

22
00:01:15,340 --> 00:01:18,270
But this is a fundamental area we
should always be on the lookout for.

23
00:01:19,310 --> 00:01:24,009
And let me close, again,
by thanking my colleague, Krishna Sridhar,

24
00:01:24,009 --> 00:01:28,246
who's really been instrumental in
the creation of the slides and

25
00:01:28,246 --> 00:01:31,421
helping with the overall
vision of this module.