0
1
00:00:00,380 --> 00:00:07,380
To create a donut chart all we have to do is add a bit more code to this cell here.
1

2
00:00:07,380 --> 00:00:13,810
So I'm going to copy it, paste it below, so I've got our original pie chart that I'm going to leave as it is.
2

3
00:00:14,220 --> 00:00:18,410
And here I'm going to add my additional code.
3

4
00:00:18,750 --> 00:00:28,020
The code I'm going to add is drawing a circle onto our chart, a white circle that is. So I'll create a variable
4

5
00:00:28,020 --> 00:00:36,630
called "center_circle" which is going to hold on to my circle and use matplotlib to
5

6
00:00:36,630 --> 00:00:43,490
create a circle, "plt.Circle", with a capital C, open parentheses,
6

7
00:00:43,650 --> 00:00:47,250
and now we have to supply some arguments.
7

8
00:00:47,250 --> 00:00:52,240
Let me hit Shift+Tab on the Circle to take a look at the quick documentation.
8

9
00:00:52,590 --> 00:01:00,420
So our Circle requires one input, one input is mandatory to create it and it basically needs to know
9

10
00:01:00,630 --> 00:01:03,410
where to center the circle.
10

11
00:01:03,480 --> 00:01:08,410
It needs an x and the y coordinate in these parentheses like so.
11

12
00:01:08,430 --> 00:01:12,540
So we're gonna give it a tuple of these coordinates.
12

13
00:01:12,540 --> 00:01:13,840
I'm just gonna center it at
13

14
00:01:14,110 --> 00:01:17,420
0, 0 and that's it.
14

15
00:01:17,990 --> 00:01:20,820
Let's add the Circle to our figure.
15

16
00:01:20,840 --> 00:01:33,710
So I'll say "plt.gca().add_artist(center_circle)".
16

17
00:01:33,710 --> 00:01:40,660
Now let me hit Shift+Enter and what we get is something like this.
17

18
00:01:40,880 --> 00:01:42,440
What's going on?
18

19
00:01:42,440 --> 00:01:51,210
Well, our circle is probably too big, it's not sized correctly, so we'll have to supply a radius.
19

20
00:01:51,320 --> 00:01:53,810
I'm going to make this circle a lot smaller,
20

21
00:01:53,870 --> 00:01:58,800
make it 0.6 for the radius and hit Shift+Enter.
21

22
00:01:58,910 --> 00:02:00,160
Much better, right?
22

23
00:02:00,290 --> 00:02:04,130
We've got the circle now sitting in the middle of our pie chart.
23

24
00:02:04,130 --> 00:02:08,570
The only problem is is that this circle is blue and it looks looks very, very odd.
24

25
00:02:08,570 --> 00:02:10,520
So let's make it white.
25

26
00:02:10,520 --> 00:02:19,160
The way we can do that is by supplying this color as an argument to a parameter called "fc", "fc = 
26

27
00:02:19,160 --> 00:02:20,720
'white'
27

28
00:02:20,930 --> 00:02:24,360
will set the color of the circle to white.
28

29
00:02:24,370 --> 00:02:27,590
Now we've got something that looks a bit more like this.
29

30
00:02:27,610 --> 00:02:30,540
This is starting to look a lot better, right?
30

31
00:02:30,580 --> 00:02:34,210
The only thing is our offset doesn't quite work anymore.
31

32
00:02:34,210 --> 00:02:37,420
It makes it look very, very strange.
32

33
00:02:37,510 --> 00:02:45,420
So I'm going to delete the explode argument here on our pie chart and refresh our graph.
33

34
00:02:45,430 --> 00:02:46,630
There we go.
34

35
00:02:46,630 --> 00:02:53,200
Not bad, but I do want to move these percent labels. Back up here on our pie chart,
35

36
00:02:53,200 --> 00:03:02,550
we can do that with an argument called "percent distance", "pctdistance", all in lower case and if I set
36

37
00:03:02,550 --> 00:03:11,280
that to 0.8, then my percent labels start moving out and they're more or less centered now
37

38
00:03:11,670 --> 00:03:14,890
in their segment of the donut chart.
38

39
00:03:14,910 --> 00:03:18,160
Now I'm fairly happy with this design.
39

40
00:03:18,210 --> 00:03:21,150
Let's review the code a little bit.
40

41
00:03:21,150 --> 00:03:25,170
We created a circle using matplotlib, so "plt.
41

42
00:03:25,200 --> 00:03:33,060
Circle()" is our equivalent of "pd.DataFrame()" and then we've supplied a couple of inputs.
42

43
00:03:33,060 --> 00:03:40,650
We've supplied a color, we've supplied a size of the circle and we've supplied where that circle should
43

44
00:03:40,650 --> 00:03:42,850
be drawn. At 0, 0
44

45
00:03:42,870 --> 00:03:49,980
it's completely centered. Then what we did is we had to add this circle to our figure.
45

46
00:03:50,700 --> 00:04:00,120
So with "plt.gca" we get the current axes, so we're getting the current axes and we're adding the circle
46

47
00:04:00,330 --> 00:04:03,980
with the "add_artist" method to the axes.
47

48
00:04:04,050 --> 00:04:09,910
So that adds the circle to our figure, which means it gets drawn on top of our pie chart.
48

49
00:04:10,620 --> 00:04:14,670
Okay, so we've almost mastered the donut chart.
49

50
00:04:14,730 --> 00:04:21,200
The only thing that I really want to show you is how to draw a donut chart with more than two categories,
50

51
00:04:21,420 --> 00:04:27,720
because oftentimes these donut charts really come into the fore when you have three, four or five categories,
51

52
00:04:28,080 --> 00:04:32,060
then they start looking really, really great in all sorts of reports.
52

53
00:04:32,130 --> 00:04:37,500
So just as an exercise, let's create a donut chart with four categories.
53

54
00:04:37,530 --> 00:04:45,520
Copy this cell, paste it below, so I'm keeping the original and working off this one and I'm going to add
54

55
00:04:45,670 --> 00:04:54,100
a few more category names, so say "Updates" and "Promotions".
55

56
00:04:54,100 --> 00:05:00,790
Now since we have four categories, I'm going to have to update my list of sizes as well.
56

57
00:05:00,790 --> 00:05:08,500
And I'm just gonna make up these numbers here for the amount of spam, legitimate emails, updates and promotions.
57

58
00:05:08,500 --> 00:05:19,250
So say we have 25 spam emails and I don't know 43 legitimate emails, 19 updates and 22 promotions.
58

59
00:05:19,480 --> 00:05:21,010
So we've got that.
59

60
00:05:21,010 --> 00:05:27,330
Now it's just a matter of picking the colors. We can keep the red and blue, but we have to add two more
60

61
00:05:27,330 --> 00:05:29,770
colors that fit well with this palette.
61

62
00:05:30,030 --> 00:05:37,500
Going back to our American Palette from Flat UI colors, I think I'm going to go for the light green and
62

63
00:05:37,500 --> 00:05:39,090
maybe the sour lemon here,
63

64
00:05:39,120 --> 00:05:40,060
the yellow.
64

65
00:05:40,060 --> 00:05:49,340
So I'm going to copy this, update our list here, making sure that the hex code of the colors and single
65

66
00:05:49,340 --> 00:05:56,360
quotes and I'm going to copy the sour lemon for the promotions.
66

67
00:05:56,380 --> 00:05:58,430
There we go.
67

68
00:05:58,430 --> 00:06:01,130
Now I've got my four categories.
68

69
00:06:01,210 --> 00:06:05,440
Let me hit Shift+Enter and see what this looks like at the moment.
69

70
00:06:05,440 --> 00:06:06,600
Not bad, right?
70

71
00:06:06,760 --> 00:06:11,540
But what if we wanted to have a small gap between each of these categories?
71

72
00:06:11,560 --> 00:06:19,290
What if we wanted to explode the donut chart the way we've exploded our pie chart? And the easiest
72

73
00:06:19,290 --> 00:06:27,690
way to do this is to create another list of values, say "offset = []" and then maybe
73

74
00:06:27,750 --> 00:06:34,730
0.05, 0.05, 0.05,...
74

75
00:06:34,770 --> 00:06:42,490
I'll make the offset equal between all of the, all of the categories, 0.05 as well.
75

76
00:06:42,880 --> 00:06:49,240
Then for the pie chart, we'll just say "explode = offset".
76

77
00:06:49,240 --> 00:06:54,180
Now let me hit Shift+Enter and look at our results.
77

78
00:06:54,180 --> 00:06:55,390
Brilliant.
78

79
00:06:55,650 --> 00:06:57,750
And that's really all there is to it.
79

80
00:06:57,810 --> 00:07:04,020
I think pie charts are, they're simple, they're clear, they're easy to code up, but you can also make them
80

81
00:07:04,020 --> 00:07:10,880
look very, very good with the right choice of colors and a little thought on the formatting. Having done
81

82
00:07:10,890 --> 00:07:14,490
some cute visualizations in the past lessons,
82

83
00:07:14,550 --> 00:07:21,090
it's now time to return to do some hardcore programming. In the next couple of lessons,
83

84
00:07:21,150 --> 00:07:24,810
we're going to be looking at natural language processing.
84

85
00:07:24,870 --> 00:07:30,510
We can't just feed raw e-mail bodies into our machine learning algorithm.
85

86
00:07:30,510 --> 00:07:36,240
We need to prepare these e-mails for our Bayes' classifier to work.
86

87
00:07:36,250 --> 00:07:41,580
I'd say this step probably falls a bit more into the data cleaning part of the project, but it will equip
87

88
00:07:41,610 --> 00:07:48,090
you with some very, very powerful tools and techniques to tackle all kinds of text data that you're gonna
88

89
00:07:48,090 --> 00:07:51,560
be encountering on your personal projects as well.
89

90
00:07:51,660 --> 00:07:57,300
Natural Language Processing is a huge topic and it's super lucrative and super interesting.
90

91
00:07:57,300 --> 00:08:00,580
So I hope you're as excited as I am for the next section.
91

92
00:08:00,660 --> 00:08:02,990
I'll see you in the next lessons.
92

93
00:08:03,000 --> 00:08:03,630
Take care.