﻿1
00:00:00,300 --> 00:00:00,600
All right.

2
00:00:00,600 --> 00:00:01,733
So that's for the train set.

3
00:00:01,733 --> 00:00:04,400
And now let's come in quickly
on the test set.

4
00:00:04,400 --> 00:00:06,900
All right.
This is already executed. Perfect.

5
00:00:06,900 --> 00:00:09,566
So here are the results
of Naive Bayes on the test set.

6
00:00:09,566 --> 00:00:13,466
And actually we realized that here
we were pretty unlucky.

7
00:00:13,500 --> 00:00:14,933
You know, naive Bayes was pretty unlucky

8
00:00:14,933 --> 00:00:18,733
because the three predictions
that it missed, you know,

9
00:00:19,066 --> 00:00:22,566
this red customer that falls incorrectly
in the green region.

10
00:00:22,566 --> 00:00:26,666
And these two green customers
that fall incorrectly in the red region.

11
00:00:26,933 --> 00:00:29,900
Well,
Naive Bayes was pretty unlucky on this one

12
00:00:29,900 --> 00:00:32,366
because, you know,
they're almost on the prediction boundary

13
00:00:32,366 --> 00:00:36,000
and it is what it missed by very closely
to beat the record.

14
00:00:36,000 --> 00:00:37,166
And same for this one actually.

15
00:00:37,166 --> 00:00:39,766
You know,
this one was very shortly missed. Right.

16
00:00:39,766 --> 00:00:43,133
Because it is an incorrect
prediction of a customer who in reality

17
00:00:43,233 --> 00:00:46,866
didn't buy the SV because it's red,
but was predicted to by DSV

18
00:00:46,866 --> 00:00:48,866
because it falls in the green region.

19
00:00:48,866 --> 00:00:50,000
So pretty unlucky.

20
00:00:50,000 --> 00:00:52,733
Which explains the accuracy of 90%.

21
00:00:52,733 --> 00:00:56,500
I think that if we tune a bit Naive Bayes,
we could beat the record.

22
00:00:56,733 --> 00:01:01,000
But once again, this will be important
and now the next step will be to buy

23
00:01:01,000 --> 00:01:03,166
the last two classification models,

24
00:01:03,166 --> 00:01:06,766
which belong to a branch of machine
learning called and Simple models.

25
00:01:07,033 --> 00:01:09,166
I'm talking,
of course, about decision tree

26
00:01:09,166 --> 00:01:12,166
classification and random forest
classification.

27
00:01:12,200 --> 00:01:15,900
And actually, with these tools
and especially random forest,

28
00:01:15,900 --> 00:01:19,166
we might have a chance to beat
that record accuracy.

29
00:01:19,166 --> 00:01:21,433
You know, we might beat 93%.

30
00:01:21,433 --> 00:01:24,466
So that's what on the menu
for this end of part three.

31
00:01:24,500 --> 00:01:26,266
So I can't wait to see you

32
00:01:26,266 --> 00:01:29,933
in the next part to build first
the decision tree classification model

33
00:01:30,066 --> 00:01:33,066
and then finally
the random forest classification.

34
00:01:33,500 --> 00:01:35,300
And until then, enjoy machine learning.
