In [1]:

import pandas as pd
from altair import Chart, X, Y, Axis, Color

Loading the dataset¶

In [2]:

outbrks_before_2014 = pd.read_csv("data/in/ebola-outbreaks-before-2014-cleaned.csv", encoding="utf-8", index_col=False)

In [3]:

outbrks_before_2014.head(2)

Out[3]:

	Start date	End date	Duration (days)	Country name	Ebola subtype	Reported number of human cases	Reported number of deaths among cases	Reported % of deaths among cases
0	2012-11-01	2013-01-31	91	Uganda	Sudan virus	6	3	0.500
1	2012-06-01	2012-11-30	182	Democratic Republic of the Congo	Bundibugyo virus	36	13	0.361

In [4]:

outbrks_before_2014.tail(2)

Out[4]:

	Start date	End date	Duration (days)	Country name	Ebola subtype	Reported number of human cases	Reported number of deaths among cases	Reported % of deaths among cases
31	1976-01-01	1976-12-31	365	Sudan (South Sudan)	Sudan virus	284	151	0.53
32	1976-01-01	1976-12-31	365	Democratic Republic of the Congo	Zaire virus	318	280	0.88

General information¶

How many outbreaks are we talking about?¶

In [5]:

"There has been {} Ebola outbreaks before 2014".format(len(outbrks_before_2014))

Out[5]:

'There has been 33 Ebola outbreaks before 2014'

Where did they happen?¶

In [6]:

num_outbrks = Chart(outbrks_before_2014).mark_bar().encode(
    X('count(*):Q', axis=Axis(
            title='Number of outbreaks',
        )),
    Y('Country name:O')  
)

num_outbrks

In [7]:

unique_countries = set(outbrks_before_2014['Country name'])
print("There has been outbreaks in {} countries:".format(len(unique_countries)))

for name in unique_countries:
    print("* in {0} -> {1}".format(name, list(outbrks_before_2014['Country name']).count(name)))

There has been outbreaks in 11 countries:
* in Uganda -> 5
* in England -> 1
* in South Africa -> 1
* in Democratic Republic of the Congo -> 9
* in Russia -> 2
* in Italy -> 1
* in Gabon -> 4
* in USA -> 3
* in Philippines -> 3
* in Sudan (South Sudan) -> 3
* in Côte d'Ivoire (Ivory Coast) -> 1

When did they happen?¶

In [50]:

outbrks_time= Chart(outbrks_before_2014).mark_bar().encode(
            x=X('Start date:T',
                axis=Axis(
                    title='time',
                ),
            ),
            x2='End date:T',
            y='Country name:N',
            color='Country name:N'
        )

outbrks_time

From the recent outbreak in West Africa we all know the severity of it.¶

But how many cases were reported and how many people deceased in previous outbreaks?¶

In [9]:

cases = Chart(outbrks_before_2014).mark_bar(color='#f2b47b').encode(
    X('average(Reported number of human cases):Q'),
    Y('Country name:O')  
)

cases

In [10]:

deceased = Chart(outbrks_before_2014).mark_bar(color='#aa4949').encode(
    X('average(Reported number of deaths among cases):Q'),
    Y('Country name:O')  
)

deceased

The two previous plots should be improved by normalising to the number of outbreak(s) in a country¶

It's hard to conclude. Although the proportions of cases/deaths in Sudan seem disproportionate compared to the number of outbreaks (3, same as USA and less than Gabon).¶

One of my main interest was to look for a link between a virus subtype and a severity of an outbreak¶

In [11]:

Chart(outbrks_before_2014).mark_circle(color='#f2b47b').encode(
    size='average(Reported number of human cases):Q',
    x=X('Ebola subtype:N'),
    y=Y('Country name:N'),
)

In [12]:

Chart(outbrks_before_2014).mark_circle(color='#aa4949').encode(
    size='average(Reported number of deaths among cases):Q',
    x=X('Ebola subtype:N'),
    y=Y('Country name:N'),
)

Once again, forgot to normalise by the number of outbreak(s). So it's hard to conclude.¶

We can just tell that the Reston virus subtype is the less severe.¶

In [52]:

cases_deaths_virus = Chart(outbrks_before_2014).mark_point().encode(
    color='Ebola subtype:N',
    x='Reported number of human cases:Q',
    y='Reported number of deaths among cases:Q',
)

cases_deaths_virus

Unsurprisingly most outbreaks show a correlation between number of cases and number of deaths.¶

Will add zoom with Vega-Lite to make it easier to spot the Reston virus seemingly leading to less cases or deaths.¶

In [14]:

Chart(outbrks_before_2014).mark_circle().encode(
    size='average(Reported number of deaths among cases):Q',
    x=X('Ebola subtype:N'),
    y=Y('Country name:N'),
    color='Country name:N'
)

Amongst the most impacted countries Sudan and Gabon have people infected by the same virus suptype, while DRC and Uganda saw two types of viruses.¶

In [15]:

df_uganda = outbrks_before_2014[outbrks_before_2014["Country name"] == "Uganda"]

In [16]:

Chart(df_uganda).mark_bar().encode(
            x=X('Start date:T',
                axis=Axis(
                    title='time',
                ),
            ),
            x2='End date:T',
            y='Ebola subtype:N',
            color='Ebola subtype:N'
)

This look at the Ugandan outbreaks remind me I should plot the number of cases/deaths and normalise by the duration of the outbreaks.¶

In [17]:

Chart(outbrks_before_2014).mark_bar().encode(
            x=X('Start date:T',
                axis=Axis(
                    title='time',
                ),
            ),
            x2='End date:T',
            y='Country name:N',
            color='Ebola subtype:N'
        )

Assuming outbreaks happening at the same period and due to the same virus subtype are due to people travelling and carrying the virus, this doesn't seem to impact nearby countries. The road networks might be better controlled than the airports.¶

In [62]:

deaths_virus_time = Chart(outbrks_before_2014).mark_circle().encode(
            size='average(Reported number of deaths among cases):Q',
            x=X('Start date:T',
                axis=Axis(
                    title='time',
                ),
            ),
            x2='End date:T',
            y='Country name:N',
            color='Ebola subtype:N'
        )

deaths_virus_time

From this dataset:¶

the countries the most impacted are DRC, Gabon, Sudan and Uganda. This makes sense as they're the ones that contain natural reservoirs of the virus hosts (bats)
the Zaire virus seems to be linked to the more outbreaks and cases/deaths, followed by the Sudan virus

Extra information:¶

the Ebola virus was first discovered in 1976
2014-2016 outbreak in West Africa (Guinea, Sierra Leone and Liberia)
this latest outbreaks was due to the Zaire virus
it was the largest and more complex and spread across borders
it led to more cases and deaths than all previous outbreaks combined:

28,637 reported cases and 11,315 deaths

sources: WHO (who.int/mediacentre/factsheets/fs103/en/), BBC (bbc.com/news/world-africa-28755033)

Let's prepare our web visualisation¶

As I aim for an interactive visualisation I'll be "cheating" here and creating json in Vega-Lite format that I'll add to an html template¶

In [59]:

import json

# Fix to add to each plot
fix = {"$schema": "https://vega.github.io/schema/vega-lite/v2.0.json",
       "description": "A simple bar chart with embedded data."}

# Add brush
brush = {"selection": {"brush": {"type": "interval"}}}

# Add paintbrush
paintbrush = {"selection": {"paintbrush": {
                              "type": "multi", "on": "mouseover",
                              "nearest": True}}}
# for a paintbrush add size into `encoding`
size = {"condition": {"selection": "paintbrush", "value": 300},
        "value": 50}

# For zoom capacity
zoom = {"selection": { "grid": {"type": "interval", "bind": "scales"}}}

In [55]:

def from_dict_to_json(data_dict, add_on, interact):
    #remove the data key/value from the plot info
    data_dict.pop("data", None)
    data_dict.update(add_on)
    data_dict.update(interact)
    return json.dumps(data_dict)

In [65]:

num_outbrks_dict = num_outbrks.to_dict()

outbrks_data_json = json.dumps(num_outbrks_dict["data"])

In [69]:

with open("data_for_viz.json", "w") as outfile:
    outfile.write(outbrks_data_json)

In [57]:

# Plot 1 - outbreaks/country
from_dict_to_json(num_outbrks_dict, fix, brush)

Out[57]:

'{"mark": "bar", "$schema": "https://vega.github.io/schema/vega-lite/v2.0.json", "encoding": {"y": {"type": "ordinal", "field": "Country name"}, "x": {"axis": {"title": "Number of outbreaks"}, "aggregate": "count", "type": "quantitative", "field": "*"}}, "description": "A simple bar chart with embedded data.", "selection": {"brush": {"type": "interval"}}}'

In [58]:

# Plot 2 - outbreaks over time
from_dict_to_json(outbrks_time.to_dict(), fix, zoom)

Out[58]:

'{"mark": "bar", "$schema": "https://vega.github.io/schema/vega-lite/v2.0.json", "encoding": {"y": {"type": "nominal", "field": "Country name"}, "color": {"type": "nominal", "field": "Country name"}, "x2": {"type": "temporal", "field": "End date"}, "x": {"axis": {"title": "time"}, "type": "temporal", "field": "Start date"}}, "description": "A simple bar chart with embedded data.", "selection": {"grid": {"bind": "scales", "type": "interval"}}}'

In [61]:

# Plot 3 - cases vs deaths per virus subtype
cases_deaths_virus_dict = cases_deaths_virus.to_dict()
cases_deaths_virus_dict["encoding"]["size"] = size
cases_deaths_virus_dict.update(paintbrush)
from_dict_to_json(cases_deaths_virus_dict, fix, zoom)

Out[61]:

'{"mark": "point", "$schema": "https://vega.github.io/schema/vega-lite/v2.0.json", "selection": {"grid": {"bind": "scales", "type": "interval"}}, "encoding": {"color": {"type": "nominal", "field": "Ebola subtype"}, "y": {"type": "quantitative", "field": "Reported number of deaths among cases"}, "x": {"type": "quantitative", "field": "Reported number of human cases"}, "size": {"value": 50, "condition": {"value": 300, "selection": "paintbrush"}}}, "description": "A simple bar chart with embedded data."}'

In [63]:

# Plot 4 - number of outbreaks
from_dict_to_json(deaths_virus_time.to_dict(), fix, zoom)

Out[63]:

'{"mark": "circle", "$schema": "https://vega.github.io/schema/vega-lite/v2.0.json", "encoding": {"y": {"type": "nominal", "field": "Country name"}, "color": {"type": "nominal", "field": "Ebola subtype"}, "x2": {"type": "temporal", "field": "End date"}, "x": {"axis": {"title": "time"}, "type": "temporal", "field": "Start date"}, "size": {"aggregate": "average", "type": "quantitative", "field": "Reported number of deaths among cases"}}, "description": "A simple bar chart with embedded data.", "selection": {"grid": {"bind": "scales", "type": "interval"}}}'

In [ ]: