﻿1
00:00:00,333 --> 00:00:00,633
Okay.

2
00:00:00,633 --> 00:00:02,800
So now we have it. So now we can use it.

3
00:00:02,800 --> 00:00:06,433
And therefore just below
we will actually call it.

4
00:00:06,700 --> 00:00:10,400
And in order to do this efficiently
we can just say the example here.

5
00:00:10,566 --> 00:00:13,600
And right below

6
00:00:13,633 --> 00:00:17,500
we paste it and we just replace once again
why true.

7
00:00:17,500 --> 00:00:19,466
But why test.

8
00:00:19,466 --> 00:00:24,066
So that we can get indeed
the accuracy under test it okay.

9
00:00:24,066 --> 00:00:25,533
And here we don't have to do a print

10
00:00:25,533 --> 00:00:30,233
because this accuracy score
will directly return that accuracy.

11
00:00:30,233 --> 00:00:32,266
You know that rate of correct predictions.

12
00:00:32,266 --> 00:00:33,066
So there you go.

13
00:00:33,066 --> 00:00:36,266
We can just play this cell
you know run this cell.

14
00:00:36,266 --> 00:00:38,333
And here you go. Congratulations.

15
00:00:38,333 --> 00:00:42,833
You have here the confusion matrix
showing that we have indeed 65

16
00:00:42,833 --> 00:00:46,833
correct predictions of the class zero,
meaning the customers of the test set

17
00:00:46,833 --> 00:00:48,533
who didn't buy the new SUV.

18
00:00:48,533 --> 00:00:53,366
Then 24 correct predictions of the class
one meaning

19
00:00:53,600 --> 00:00:56,966
correct predictions of the customers
who but the SUV

20
00:00:57,300 --> 00:01:01,033
and then three incorrect
predictions of the class one

21
00:01:01,033 --> 00:01:04,033
meaning three incorrect
predictions of the customers who

22
00:01:04,166 --> 00:01:07,666
but in reality the SUV but were predicted
not two.

23
00:01:08,033 --> 00:01:13,000
And finally eight incorrect
predictions of the class zero mean

24
00:01:13,333 --> 00:01:16,433
eight customers who in reality didn't buy

25
00:01:16,433 --> 00:01:19,433
the SUV but were predicted to buy it.

26
00:01:19,533 --> 00:01:23,000
Okay, so you see, the confusion
matrix has no mystery.

27
00:01:23,000 --> 00:01:26,033
It's very easy to read,
and in a flashlight we can indeed

28
00:01:26,033 --> 00:01:29,200
get the main information
of our predictions.

29
00:01:29,800 --> 00:01:34,100
And finally, that little number that
we have here is of course the accuracy.

30
00:01:34,366 --> 00:01:37,566
And here we get oh point 89,

31
00:01:37,566 --> 00:01:42,400
which means that we had 89%
of correct predictions in the test set.

32
00:01:42,600 --> 00:01:46,400
And actually remember that
there are 100 observations in a test set,

33
00:01:46,566 --> 00:01:50,400
which means that we had indeed
89 correct predictions.

34
00:01:50,400 --> 00:01:54,833
Actually, 65 here
plus 24 is equal indeed to 89.

35
00:01:54,966 --> 00:01:55,333
All right.

36
00:01:55,333 --> 00:01:58,433
But for any size of the test set,
this would mean that

37
00:01:58,433 --> 00:02:01,433
you had 89% of correct predictions.

38
00:02:01,433 --> 00:02:03,300
And that's exactly the accuracy.

39
00:02:03,300 --> 00:02:06,433
The accuracy is the rate
of correct predictions.

40
00:02:07,166 --> 00:02:07,600
All right.

41
00:02:07,600 --> 00:02:12,066
So now you know how to, you know, quickly
evaluate a classification model.

42
00:02:12,066 --> 00:02:13,866
You know the accuracy is usually

43
00:02:13,866 --> 00:02:17,900
the right metric to use when evaluating
your classification models.

44
00:02:18,066 --> 00:02:20,133
So now you have it in the tool kit.

45
00:02:20,133 --> 00:02:23,766
And so here we go
for our final step of the journey.

46
00:02:23,933 --> 00:02:26,933
We're going to visualize
not only the training

47
00:02:26,933 --> 00:02:30,000
set results but also the test set results.

48
00:02:30,000 --> 00:02:33,766
And this will be super interesting
because we will actually see

49
00:02:34,033 --> 00:02:36,800
how the logistic regression classifier

50
00:02:36,800 --> 00:02:40,200
was actually trained
to classify our customers.

51
00:02:40,200 --> 00:02:45,366
You know, our observations into the two
different classes, you know, 0 or 1.

52
00:02:45,766 --> 00:02:49,433
We will have super nice results
showing all the real results

53
00:02:49,433 --> 00:02:51,000
in both the training set and the test.

54
00:02:51,000 --> 00:02:55,333
It and also the prediction regions,
you know, the regions where

55
00:02:55,466 --> 00:02:59,466
our predictions are zero and the other
region where the predictions are one.

56
00:02:59,800 --> 00:03:03,000
And you will see that the curve
that separates

57
00:03:03,000 --> 00:03:06,166
these two regions is exactly
the classification curve.

58
00:03:06,166 --> 00:03:09,866
And you will see that it will be
a straight line for linear models

59
00:03:09,966 --> 00:03:13,066
and something else
in a straight line for nonlinear models.

60
00:03:13,066 --> 00:03:15,600
I really, really
can't wait to show you this

61
00:03:15,600 --> 00:03:19,666
because you will really see and visualize
the difference between

62
00:03:19,666 --> 00:03:23,466
linear classification models
and nonlinear classification models.

63
00:03:23,833 --> 00:03:24,600
So there you go.

64
00:03:24,600 --> 00:03:27,466
Take a little break
now, and we'll tackle together

65
00:03:27,466 --> 00:03:30,933
this final step to visualize
indeed these two results.

66
00:03:30,933 --> 00:03:32,900
And until then, enjoy machine learning.
