0
1
00:00:00,700 --> 00:00:07,350
All right, so in the upcoming lessons I want to talk a little bit more about data visualization. In particular,
1

2
00:00:07,430 --> 00:00:11,040
I want to to show you guys how to create some beautiful pie charts.
2

3
00:00:11,150 --> 00:00:18,690
And afterwards I wanna show you how to create a variation on the pie chart called the doughnut chart.
3

4
00:00:18,690 --> 00:00:23,350
Now, both of these types of charts are very, very common, especially in business reports,
4

5
00:00:23,520 --> 00:00:28,830
so it's very important that we also get these two charts under our belt.
5

6
00:00:28,850 --> 00:00:37,250
The first thing I'll do of course is add another markdown cell and I'll call this "Number of Spam Messages
6

7
00:00:37,820 --> 00:00:41,340
Visualised (
7

8
00:00:41,340 --> 00:00:44,640
Pie Charts)". There we go.
8

9
00:00:44,650 --> 00:00:50,350
Now before we dive into the visualizations, we should probably check how many spam and how many non-spam
9

10
00:00:50,350 --> 00:00:53,720
messages there are. For this, our dataframe's
10

11
00:00:53,830 --> 00:00:56,810
value_counts method is going to be very, very handy.
11

12
00:00:57,160 --> 00:01:08,710
So if we take our dataframe and select the CATEGORY column, we can use value_counts and then
12

13
00:01:08,710 --> 00:01:15,140
parentheses to get a breakdown of how many messages of each category we've got.
13

14
00:01:15,370 --> 00:01:21,100
Here we can see that we've got 3900 non-spam or ham messages and we've got 
14

15
00:01:21,100 --> 00:01:25,680
1896 spam messages.
15

16
00:01:25,780 --> 00:01:30,550
Now, since we're gonna be using these numbers later on, let's store them in some variables.
16

17
00:01:30,730 --> 00:01:33,580
So I'll say "amount_of_
17

18
00:01:33,580 --> 00:01:37,310
spam = data.
18

19
00:01:37,420 --> 00:01:39,890
CATEGORY.value_counts()[
19

20
00:01:40,140 --> 00:01:42,770
1]".
20

21
00:01:43,410 --> 00:01:52,620
And the amount of ham is gonna be equal to the same thing except it's going to be followed by "[0]".
21

22
00:01:52,620 --> 00:02:02,100
Zero because I access the first value in this series and one to access the second value
22

23
00:02:02,400 --> 00:02:04,310
in this series.
23

24
00:02:04,380 --> 00:02:05,970
Now, to do our visualization,
24

25
00:02:05,990 --> 00:02:09,180
we're gonna be using our old friend matplotlib.
25

26
00:02:09,180 --> 00:02:17,250
So let's import matplotlib's pyplot functionality as plt at the top of our notebook. Going to our
26

27
00:02:17,250 --> 00:02:19,770
notebook imports we'll say "import
27

28
00:02:22,590 --> 00:02:30,940
matplotlib.pyplot as plt".
28

29
00:02:30,940 --> 00:02:33,430
Now, as usual, we're also gonna be adding
29

30
00:02:33,560 --> 00:02:38,210
"%matplotlib inline".
30

31
00:02:38,450 --> 00:02:42,010
This line of code here is Jupyter notebook specific
31

32
00:02:42,230 --> 00:02:50,540
and it allows us to export our images for our charts when we export the notebook.
32

33
00:02:50,570 --> 00:02:58,310
So if you ever go to Download, Download notebook as Notebook, then if you have this line of code in here
33

34
00:02:58,640 --> 00:03:00,530
then you get the graphics as well.
34

35
00:03:01,410 --> 00:03:01,890
All right.
35

36
00:03:01,910 --> 00:03:04,730
So how do you create one of these friendly pie charts?
36

37
00:03:04,760 --> 00:03:12,160
The first thing I'll do is I'll create a list of our category names, so category_names
37

38
00:03:12,200 --> 00:03:15,950
is gonna be equal to a list with the square brackets,
38

39
00:03:15,950 --> 00:03:27,340
"['Spam', 'Legit Mail']" and the next thing I'll do is create
39

40
00:03:27,430 --> 00:03:28,480
another list.
40

41
00:03:28,720 --> 00:03:34,240
I'm going to call this one "sizes" for the size of the different pieces of the pie.
41

42
00:03:35,430 --> 00:03:45,180
This is going to hold on to the amount_of_spam and it's going to hold on to the amount_of_ham.
42

43
00:03:45,210 --> 00:03:57,450
Now we can use these two to generate our chart, "plt.pie(sizes, " and then the labels
43

44
00:03:57,930 --> 00:04:08,660
of the chart are gonna be equal to the category_names. With "plt.show()" we can see what our basic,
44

45
00:04:08,810 --> 00:04:12,020
very, very basic pie chart is going to look like. Now,
45

46
00:04:12,050 --> 00:04:17,060
I don't recall if I actually pressed Shift+Enter a minute earlier on my notebook cell, but let me hit
46

47
00:04:17,060 --> 00:04:20,510
Shift+Enter now and we'll quickly find out.
47

48
00:04:21,480 --> 00:04:25,490
Nope, get my "NameError: 'plt' is not defined".
48

49
00:04:25,490 --> 00:04:27,140
This means that we have to go back up,
49

50
00:04:27,790 --> 00:04:32,090
or I do at least, and hit Shift+Enter on my imports.
50

51
00:04:32,090 --> 00:04:33,300
There we go.
51

52
00:04:33,300 --> 00:04:41,940
Now I can come down here and actually generate my pie chart and it looks like this. It's pretty ugly actually.
52

53
00:04:42,470 --> 00:04:47,090
So we're gonna have to customize this and make it look presentable.
53

54
00:04:47,090 --> 00:04:48,530
We're going to dress it up.
54

55
00:04:48,630 --> 00:04:54,810
The first thing I'm going to do is actually make those font sizes on this thing look a bit larger, so I can access
55

56
00:04:54,960 --> 00:05:01,360
the font sizes of my labels here with the property called "textprops";
56

57
00:05:01,620 --> 00:05:06,100
"textprops = {
57

58
00:05:06,180 --> 00:05:12,540
'fontsize': }", say, I don't know, maybe 6,
58

59
00:05:12,600 --> 00:05:13,340
see what happens.
59

60
00:05:14,400 --> 00:05:16,210
OK, they get even smaller,
60

61
00:05:16,210 --> 00:05:17,210
not what I want.
61

62
00:05:17,740 --> 00:05:20,200
Maybe I'll pick 16.
62

63
00:05:20,200 --> 00:05:21,520
There we go.
63

64
00:05:21,520 --> 00:05:27,160
That's a bit better, but this whole thing is still very, very unconvincing.
64

65
00:05:27,160 --> 00:05:31,360
For starters I can see that it's not very, very sharp on my screen.
65

66
00:05:31,360 --> 00:05:36,110
So what I can do in this case is manipulate the figure itself, so "plt.
66

67
00:05:36,120 --> 00:05:43,760
figure(figsize = )",
67

68
00:05:43,760 --> 00:05:48,220
I don't know, say, "2,2", right.
68

69
00:05:48,410 --> 00:05:50,540
Changing the figure size.
69

70
00:05:50,660 --> 00:05:51,840
See what this looks like.
70

71
00:05:51,950 --> 00:05:57,740
It gets a lot smaller, but this isn't the only thing I can change on the figure size.
71

72
00:05:58,370 --> 00:06:05,270
If I wanted to make this whole thing look tack sharp on my screen, then I can actually set the density
72

73
00:06:05,900 --> 00:06:11,960
of the pixels per inch or DPI to whatever my monitor supports.
73

74
00:06:11,960 --> 00:06:15,250
So I've got a pretty decent monitor that I'm working with here.
74

75
00:06:15,350 --> 00:06:22,480
It has quite a high resolution, so it actually supports 227 pixels per inch.
75

76
00:06:22,700 --> 00:06:29,030
And this will have an interesting effect, so if I press Shift+Enter now, you can see that the whole thing
76

77
00:06:29,630 --> 00:06:35,780
starts looking first of all a lot larger because it's scaled up and second of all the edges start looking
77

78
00:06:35,780 --> 00:06:37,980
a bit more clear.
78

79
00:06:38,010 --> 00:06:43,980
Only thing is my font size is probably a bit too large now, so if I go back to font size say 6,
79

80
00:06:44,120 --> 00:06:48,210
then it starts looking a bit better like this.
80

81
00:06:48,330 --> 00:06:55,920
Now, when you're working on this project, have a play with the different figure sizes, the DPI values and
81

82
00:06:55,920 --> 00:06:58,910
the font size and see how it affects your scaling.
82

83
00:06:58,950 --> 00:07:04,080
You're probably gonna have to use a bit of trial and error to get this thing looking good or looking
83

84
00:07:04,080 --> 00:07:06,640
the way you want it to on your monitor.
84

85
00:07:07,560 --> 00:07:12,540
If you ever want to export this thing and you know save it to, put it into a report or what have you,
85

86
00:07:12,540 --> 00:07:20,070
it's probably better to create a larger version of this figure and then just right click and save the
86

87
00:07:20,070 --> 00:07:26,670
image as a larger image, because then if you ever need to include it into a report or into a Word document
87

88
00:07:26,730 --> 00:07:33,540
or whatever you're using, then you can downscale it and it'll look a lot better than taking a small image and
88

89
00:07:33,540 --> 00:07:35,350
scaling it up.
89

90
00:07:35,350 --> 00:07:39,980
Now let me pull up the quick documentation on the "pie" functionality from matplotlib.
90

91
00:07:40,800 --> 00:07:46,910
We can actually see that there's quite a few different parameters that we can set.
91

92
00:07:46,920 --> 00:07:48,320
Let me show you two of them.
92

93
00:07:48,510 --> 00:07:56,730
I want to show you the start angle which is currently set to none and I want to show you this auto percent
93

94
00:07:56,820 --> 00:08:00,450
parameter that we've got here and how to use it.
94

95
00:08:00,480 --> 00:08:04,500
Let's have a play with the start angle first.
95

96
00:08:04,500 --> 00:08:08,370
So if I set the start angle to zero, let's see what we have.
96

97
00:08:08,730 --> 00:08:16,110
We can see that the chart looks like so, then we can see that this blue portion just rotated by 10 degrees
97

98
00:08:16,350 --> 00:08:18,380
from this position.
98

99
00:08:18,630 --> 00:08:28,380
You can see this rotation actually very clearly if I start going to 20, 30, 40, 50 and so on. The blue piece
99

100
00:08:28,590 --> 00:08:33,390
starts rotating counterclockwise from the x axis,
100

101
00:08:33,390 --> 00:08:39,470
so from a straight line that goes from the center out to here.
101

102
00:08:39,510 --> 00:08:49,380
If I set the start angle to 90, then our chart rotates by exactly one fourth of the way around counterclockwise.
102

103
00:08:49,380 --> 00:08:49,790
All right.
103

104
00:08:49,820 --> 00:08:51,560
So that's the start angle.
104

105
00:08:51,600 --> 00:08:55,840
Now let's have a look at the other property that I wanted to show you -
105

106
00:08:56,090 --> 00:08:58,300
auto percent,
106

107
00:08:58,350 --> 00:09:09,650
and we're gonna set that equal to '%1.1f%%' and
107

108
00:09:09,650 --> 00:09:19,220
what we get now is we get a pie chart where the percent of the chart is displayed on the chart itself.
108

109
00:09:19,220 --> 00:09:24,070
The formatting is to one decimal point, because we've put 1.1 here.
109

110
00:09:24,290 --> 00:09:31,250
If instead I had "1.2f", then it would show me two decimal points.
110

111
00:09:31,250 --> 00:09:36,890
So depending on the amount of precision that you want on your pie chart you can choose the formatting
111

112
00:09:37,040 --> 00:09:40,150
of the percent sign as appropriate.
112

113
00:09:40,450 --> 00:09:40,740
Now,
113

114
00:09:40,740 --> 00:09:46,880
personally, I don't like showing too many digits after the decimal point. In most use cases,
114

115
00:09:46,940 --> 00:09:51,990
you usually want to favor readability over precision when it comes to these visualizations anyhow.
115

116
00:09:53,200 --> 00:09:58,930
If we wanted to round this to the nearest percentage, we can put a zero here and then we'll get 33 and
116

117
00:09:58,930 --> 00:10:00,810
67.
117

118
00:10:00,850 --> 00:10:03,250
Now I find this chart here,
118

119
00:10:03,250 --> 00:10:07,370
size wise and formatting wise, is starting to look pretty decent.
119

120
00:10:07,390 --> 00:10:09,050
The only thing that I don't like,
120

121
00:10:09,050 --> 00:10:13,410
are these really awful colors that it's showing me here by default.
121

122
00:10:13,450 --> 00:10:18,220
So let's, let's change this and try to improve the design a little bit.
122

123
00:10:19,400 --> 00:10:21,760
Just so we have a record of our old chart,
123

124
00:10:21,830 --> 00:10:25,340
I'm going to copy this cell and I'm going to paste it below.
124

125
00:10:25,340 --> 00:10:32,000
Now I've got two different pie charts and I'm going to spice up the design on this one and leave the
125

126
00:10:32,000 --> 00:10:33,580
other one as it is.
126

127
00:10:33,950 --> 00:10:40,330
One of my go to places for picking a nice palette is a website called Flat UI Colors.
127

128
00:10:40,400 --> 00:10:46,170
Let's check out this American Palette here that they've got. The first color that I'm gonna grab is gonna
128

129
00:10:46,190 --> 00:10:57,230
be maybe Pink Glamour, means I can go back here into my Jupyter notebook and save my colors as a list.
129

130
00:10:57,230 --> 00:11:06,890
I'm going to call it a "custom_colors = ['']", paste. This is gonna be my first
130

131
00:11:06,890 --> 00:11:13,880
color, comma, back to the American Palette and I'm going to select the color that goes well with this
131

132
00:11:13,880 --> 00:11:14,970
other category.
132

133
00:11:15,080 --> 00:11:19,320
So I think something that would be contrast-y would be this blue one here.
133

134
00:11:19,510 --> 00:11:22,660
So I'm gonna take this one, go back here,
134

135
00:11:24,430 --> 00:11:32,860
add that as well. Having added my colors' hex codes to a list, I can now feed them in as an argument to
135

136
00:11:32,890 --> 00:11:34,420
my "pie" method.
136

137
00:11:34,960 --> 00:11:47,980
So, "colors =custom_colours" and Shift+Enter will color in my pie chart in a much more beautiful way
137

138
00:11:48,660 --> 00:11:57,400
and I've picked some light colors, so that the dark text, this 33 and 67 percent is still very, very readable.
138

139
00:11:57,400 --> 00:12:03,340
Now one thing we can also do to make the design a little bit snappier is to break out this spam section
139

140
00:12:03,340 --> 00:12:10,990
here from the legitimate e-mail section, so we can actually have a little bit of a gap in between the
140

141
00:12:10,990 --> 00:12:21,240
two sections and we can do this by supplying an argument here called "explode", so "explode = "
141

142
00:12:22,170 --> 00:12:25,150
and then we also supply a list.
142

143
00:12:25,230 --> 00:12:31,800
This list will have two values, say 0 and 0.1.
143

144
00:12:32,130 --> 00:12:36,300
These numerical values here set the size of the gap.
144

145
00:12:36,420 --> 00:12:44,230
So let me press Shift+Enter and show you what this would look like. So passing in an argument of 0
145

146
00:12:44,350 --> 00:12:49,150
and 0.1 will get the chart looking like this.
146

147
00:12:49,150 --> 00:12:55,190
If we had 0.5 instead, then the gap would start looking a lot bigger.
147

148
00:12:55,750 --> 00:13:02,860
So I'm going to set this back to 0.1 and I can also meddle with the first one here, so 0.5
148

149
00:13:02,890 --> 00:13:09,760
instead of 0 and we get again something like this. The offset that we're applying here,
149

150
00:13:10,300 --> 00:13:18,220
if we go to the quick documentation essentially specifies the fraction of the radius with which to offset
150

151
00:13:18,520 --> 00:13:27,380
each wedge. Since we have two wedges, the red one and the blue one, we're supplying two numbers. So in my
151

152
00:13:27,380 --> 00:13:31,250
case, I'm offsetting one of them by 0.1.
152

153
00:13:31,640 --> 00:13:38,560
If I offset both of them by 0.1, then the gap between the two wedges doubles.
153

154
00:13:38,580 --> 00:13:42,840
Now I quite like the look of a small offset, 0.1 will do for me.
154

155
00:13:43,950 --> 00:13:51,280
In the next lesson I'll show you how to take these designs a little further and make a donut chart. I'll
155

156
00:13:51,280 --> 00:13:52,140
see you there.