Preview
Video Player
00:00
00:00
Plotting the World
16:35Charts and graphs are definitely a step up from viewing data in a spreadsheet. Let's take a look at how Bokeh can take visualizations even further.
Now that we have seen how we can plot
data to see if there's potentially
0:00
a correlation between two factors,
like population and life expectancy.
0:04
It might be interesting to see if we can
visualize our data in a different way,
0:09
something beyond the normal bar chart,
or scatter plot.
0:13
First though, now that we have
a basic understanding of Bokeh,
0:17
let's take a slight step back and
0:21
see a few other features that Bokeh
provides in order to see if we can
0:23
generate a potentially more interesting
visualization of a population data.
0:27
We have been using shape glyphs for
0:31
our scatter plot, which have been
very relevant for our work this far.
0:33
Another method we can call in our
figure object is called patches,
0:38
which allows us to draw shapes
based on passed-in coordinates.
0:42
We can then use the fill color parameter
to set the color of our shape.
0:47
Let's start with some basic
shapes to see how this works,
0:52
along with some additional parameters
we can use and see if we can't generate
0:55
something that might be interesting
with our raw population data.
0:59
To get started,
let's go into a new file called shapes.py
1:03
and set up our imports
from bokeh.plotting,
1:08
we'll want figure, output_file and show.
1:11
Now we need to name our output file.
1:17
Lets call it shapes.html.
1:19
Next, we need to create our figure and
assign it a name.
1:22
For now, let's just call it plot.
1:25
We'll make the width and
height 400 pixels.
1:28
And we'll give it a title called Shape.
1:39
Great, now we're wanting to draw a shape.
1:42
For a single shape,
we will use the patch method.
1:46
But we'll see the patches
method here shortly as well.
1:49
We provide a list of x values to use along
with an equal length list of y values into
1:52
the patch method, which results in that x,
y coordinate pairing again.
1:57
Then let's assign a color and
alpha value and align width of two pixels.
2:01
Patch do 1, 2, 3 and 4 for x values,
try 7, 12, 9, and 3 for our y values.
2:08
For our color, 2B5B84 for
2:18
the hex value, 0.7.
2:25
line_width = 2,
which looks like it puts us over
2:32
our PEP-8 for line length so
we'll knock that down there.
2:35
That's looking good.
2:41
Now we add our method to show our plot.
2:43
And run our script.
2:47
We see that Bokeh generates our
shapes.html output file and
2:53
displays it in our browser thanks Bokeh.
2:57
We should see a four sided polygon,
3:00
filled with a slightly transparent
blue color displayed in our browser.
3:03
But what if we want to draw
multiple shapes on the same figure?
3:06
That's where patches comes in.
3:10
As the names would indicate,
we will use patch for single polygons and
3:12
patches for displaying multiple polygons.
3:16
Let's generate another plot.
3:19
plot_multiple.
3:24
We'll use the same plot width and height.
3:24
And we'll call it Multiple Shapes.
3:36
Now, to generate multiple shapes,
3:43
we need to pass in multiple arguments
into our parameters in the form of list.
3:45
For our coordinates,
we pass in a list of the x values for
3:50
our shapes, then a list of our y values.
3:53
Again, the list of the x and
3:56
y coordinates have to be the same
length overall for each shape itself.
3:58
Since we'll be generating multiple
polygons, we will want the patches method,
4:02
plot_multiple.patches.
4:07
As values for our first shape,
we'll use [1, 1,
4:12
4, 4], [3, 5, 9] for a second shape.
4:17
y values [1, 4, 4, 1], [1,
4:22
9, 3] for our second shape.
4:27
It should come as no surprise here.
4:31
They will need to have list of
the same length for our color, alpha,
4:33
and line_width values as well.
4:37
Hash first color, for our second color,
4:39
First alpha and the second alpha.
4:55
And line widths of 2 and 3.
5:03
Let's display this new plot at that
same time next to the previous plot.
5:05
For that,
we'll pass in row into the show method.
5:09
And we can't forget to import row.
5:23
From bokeh.layout import row.
5:33
Let's run our script and take a look.
5:38
There, we have our two plots.
5:43
One with a single blue tone polygon and
the latest one with a green square and
5:45
overlapping gray triangle.
5:50
Now that we have some experience
with plotting shapes,
5:52
let's get into drawing something
a little more complex.
5:55
We should be able to use longitude and
latitude coordinates as our x and
5:58
y values to generate a world map.
6:02
I have included a link to a JSON
resource that has longitude and
6:05
latitude coordinates for most of the
countries and land masses of the world.
6:09
We'll read them the coordinates as x and
y values for
6:13
our patches plot and
then have bokeh draw our map.
6:16
Let's move into a new file
called the world_map.py.
6:20
To get started, we will want to import,
request, and numpy.
6:24
We'll use request to get our JSON data,
6:28
and we'll want to bring it
on our Bokeh import as well.
6:30
We'll need output_file and show,
and since we'll be using HoverTool,
6:32
column data source and figure,
we'll want to import those as well.
6:36
ColumnDataSource and figure.
6:46
We'll name our output_file
world-map.html and
6:52
we'll give it a title of World Map. We'll
6:58
create a URL convenience variable for
our JSON string.
7:05
Again, check the teacher's notes.
7:10
We'll be using the same
tools we have been using.
7:13
So let's make a TOOLS variable for
those two.
7:16
Pan, wheel_zoom, box_zoom,
7:21
reset, hover and save.
7:27
Next, we need to use request to
get the JSON data from our URL,
7:33
request.get our url and our json_data
7:39
We get our json_data.
7:46
Okay, this data set is a little
interesting in that the coordinates for
7:48
each country can be represented at
different depths in the JSON object,
7:52
dependent on the continuity
of the country's borders.
7:57
If a country is a single
polygon such as Uganda,
8:00
it is represented in a single
list of coordinates.
8:04
If however, it is a multi-polygon, for
example, the Philippines with multiple
8:07
different land masses, it is represented
in a multi-depth list in the JSON data.
8:12
I have provided a function that parses
through these different cases and
8:17
generates the required x and
y pairs for Bokeh to use.
8:21
Check the teacher's notes for
a copy of the code.
8:25
Now we can get our coordinates and
set up our plot.
8:28
We'll configure our plot,
call this plot world_map_plot.
8:45
We'll make the plot width a little wider
here to make the map look correct.
8:52
Our tools will be our TOOLS.
9:11
x_range of the graph will be -180 to 180.
9:15
y_range -90 to 90.
9:22
There we go.
9:28
Now we assign it the patches method.
9:31
For our x values we'll use lats,
for y values we'll use longs.
9:35
Our fill_color,
9:42
let's use #F1EEF6,
9:46
fill_alpha 0.7 and
9:50
align_width will be 2.
9:55
Now we can show our plot.
10:04
And run.
10:10
We are presented with a nice
visualization of a map of the world.
10:14
We can definitely use this map for
our world population data visualization.
10:17
We'll need to convert our
country data with latitudes and
10:22
longitudes into a pandas data frame,
10:25
merge it with our population data,
and use that as a column data source.
10:29
Let's create an empty list to
handle our country geospacial data.
10:33
country_coords, and it populates
10:36
that list with our information.
10:42
land_mass.
10:58
country_code is gonna be country.
11:04
"longitudes": longs
11:21
We'll append land_mass to
our country coordinates.
11:36
Then, we can convert that into
a pandas data frame to prepare it for
11:44
an upcoming merge.
11:48
But we didn't import pandas.
11:58
So let's go back up and do that.
12:00
There we go.
12:07
All right, next, we bring in our
population data we were working with in
12:11
an earlier stage, as a data frame.
12:15
input_file, these are country-pops.csv.
12:20
Country populations.
12:33
There we go.
12:40
And then we merge the two datasets
together using pandas merge function for
12:42
use as a column data source.
12:47
We can merge on the three letter
ISO value from country-pops.csv.
12:49
And then country code value from the JSON
file in an outer join cell merge.
12:54
Three from our country codes.
13:09
country_code and how is outer.
13:17
With our merged data,
we can create a column data source.
13:27
Then, we can figure our patches plot
to utilize our column data source
13:41
information.
13:46
So we'll replace our x values here.
13:48
And then add in our source,
as our map data.
14:08
Finally, we generate our hover tools.
14:14
In our hover tools, we'll keep
the same data as we've seen before,
14:17
country name, population,
and life expectancy.
14:21
Feel free to add in additional
data to the tool tips on your own.
14:25
We want country name in English to
map to our Country_English, right?
14:56
Population.
15:12
Map to Population.
15:19
life_expectancy in years.
15:23
To map to life expectancy.
15:32
Let's try it out.
15:39
Awesome.
15:43
Now when we hover over a country, we get
our data included with the hover tool.
15:44
We can examine the different
countries on the map and
15:48
visually see, in an interactive fashion,
what the data is for each country.
15:52
Further, since we are all generally
familiar with the world map,
15:57
it provides much better context for
16:01
us about where population centers
are on a worldwide level.
16:03
That is in my opinion a pretty nice
16:07
upgrade to visualizing our world
population data in a spreadsheet.
16:10
This is just the beginning of your
journey with data visualization though.
16:16
There are many, many different
ways to explore datasets and
16:19
we have just scratched the surface for
what Bokeh can do.
16:22
Leave a message in the community forum or
contact me on Twitter and
16:25
let me know how you have used
Bokeh to better understand data.
16:29
I'll see you next time, and happy coding.
16:32