Abdel An’lah Tidjani - Political and Conflict Event in Ghana

Exploratory Data Analysis using using Python’s Pandas library to understand the data, identify partterns of conflict and making initial observations. Matplotlib and Seaborn libraries for simple visualization. Plotly library for interactive visualisation.

Notebook configuration

Code

import warnings

warnings.simplefilter(action='ignore', category=FutureWarning)

import numpy as np
import pandas as pd
from datetime import datetime
import seaborn as sns
import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import altair as alt
from IPython.display import Markdown
from tabulate import tabulate

from Wrangledata import wrangle #Our function

# Number of rows and columns
pd.set_option('display.max_rows', 150)
pd.set_option('display.max_columns', 150)
pd.set_option('display.max_colwidth', 500)

Data

Code

# Load data
dataset_name = 'data/1980-11-04-2021-11-13-Ghana.csv'
df = wrangle(dataset_name)

Inspect the Structure

The data frame contains 18 attributes, and 1,276 observations.

Code

print("Shape of the data:", df.shape)

Shape of the data: (1276, 18)

Code

Rows,Cols=df.shape

# Data columns
columns = df.columns
print("Data columns :\n",columns)

# show first 5 columns and rows

Data columns :
 Index(['event_date', 'year', 'event_type', 'sub_event_type', 'actor1',
       'inter1', 'actor2', 'inter2', 'interaction', 'admin1', 'admin2',
       'location', 'latitude', 'longitude', 'source', 'source_scale', 'notes',
       'fatalities'],
      dtype='object')

Code

Markdown(tabulate( 
    df.head()[columns[:5]],
    headers=columns[:5],
    showindex=False
))

event_date	year	event_type	sub_event_type	actor1
2021-11-04 00:00:00	2021	Protests	Peaceful protest	Protesters (Ghana)
2021-11-03 00:00:00	2021	Riots	Violent demonstration	Rioters (Ghana)
2021-11-03 00:00:00	2021	Protests	Peaceful protest	Protesters (Ghana)
2021-11-01 00:00:00	2021	Riots	Mob violence	Rioters (Ghana)
2021-10-30 00:00:00	2021	Battles	Armed clash	Nadu Warriors Communal Militia (Ghana)

Map of conflict event in Ghana

Code



# Plot Mapbox location and event type

fig = px.scatter_mapbox(
    df, # Our DataFrame
    lat="latitude",
    lon="longitude",
    # width=1000, # Width of map
    # height=700, # Height of map
    # size="fatalities",
    color="event_type",
    # Display notes when hovering mouse over event
    hover_data=["interaction","year","fatalities","location"], 
)
fig.update_layout(mapbox_style="open-street-map")

Data exploratory

Event

Code

# number of records

print("event county count:",df["admin1"].nunique(),"\n",
    "event state count:",df["admin2"].nunique(),"\n",
    "event type count:",df["event_type"].nunique(),"\n",
    "event sub_event_type count:",df["sub_event_type"].nunique())

event county count: 16 
 event state count: 175 
 event type count: 5 
 event sub_event_type count: 16

Code

event_type=df["event_type"].value_counts()
event_type


# Interactive with plotly

Protests                      486
Riots                         343
Violence against civilians    246
Battles                       165
Strategic developments         36
Name: event_type, dtype: int64

Code

fig = go.Figure(data=[go.Pie(labels=event_type.index, 
                             values=event_type, 
                             pull=[0.1, 0, 0.2,0, 0])])
fig.update_layout(
    title_text="<b> Total Type of Event percentage since 1997<b>",
    titlefont={'color':None, 'size': 15, 'family': 'San-Serif'},
    # showlegend=False, 
    # autosize=False,
    # width=600,
    # height=400,
    )

# event and sub_event

Code

sub_event=df.groupby(["event_type","sub_event_type"]
            )["event_date"].agg({'count'}).reset_index()

# sub_event

# Event type and Sub event type visualisation 

fig = (px.sunburst(sub_event, 
                  path=['event_type', 'sub_event_type'], 
                  values='count',
                 width=600, # Width of map
                height=600,)
.update_traces(textinfo="label+percent parent")
.update_layout(
    title="<b> Event type and Sub event type percentage in Ghana (ACLED)<b>",
    titlefont={'color':None, 'size': 15, 'family': 'San-Serif'},
    # showlegend=False, 
    # height=500, 
    # width=750,
    )
)
fig

Actor type involved

Code

actor_type1=df["inter1"].value_counts()
actor_type1

protesters            483
rioters               339
communal militias     173
political militias    167
state forces           89
other forces           19
rebel groups            6
Name: inter1, dtype: int64

Code

actor_type2=df["inter2"].value_counts()
actor_type2

no victicme           503
civilians             357
state forces          201
communal militias     104
rioters                68
political militias     35
other forces            3
protesters              3
rebel groups            2
Name: inter2, dtype: int64

Code

fig = (make_subplots(rows=1, cols=2, specs=[[{'type':'domain'}, {'type':'domain'}]])
.add_trace(go.Pie(labels=actor_type1.index, values=actor_type1,name="Actor type1"),1,1)
.add_trace(go.Pie(labels=actor_type2.index, values=actor_type2,name="Actor type2"),1,2)
.update_traces(hole=.6, hoverinfo="label+percent+name") 
.update_layout(
    title="<b> Actor type involved in Ghana conflict<b>",
    titlefont={'color':None, 'size': 20, 'family': 'San-Serif'},
    # showlegend=False, 
    # height=600, 
    # width=1000,
    # Add annotations in the center of the donut pies.
    annotations=[dict(text='<b>instigator', x=0.16, y=0.5,
                font_size=10, showarrow=False),
                dict(text='<b>victime/hit back', x=0.88,
                 y=0.5, font_size=10, showarrow=False)]
    )
)
fig

Conflict trend

Let look at the trend of conflict

Code

from matplotlib.dates import DateFormatter
import matplotlib.dates as mdates

# Add labels to the plot
style = dict(size=12, color='black')

fig, ax = plt.subplots(figsize=(17, 7))

number_ev=df["event_type"].groupby([df['event_date'].dt.year.rename('year'), 
             df['event_date'].dt.month.rename('month')]).agg({'count'})

number_ev.index = [pd.datetime(year, month,1)
                   for (year,month) in number_ev.index]

number_ev.reset_index(inplace=True)

ax.plot('index','count', data=number_ev,color="b")

# Format the x axis 
ax.xaxis.set_major_locator(mdates.YearLocator(base=1))

ax.grid(True)
plt.ylabel("Number of conflict event")
plt.title("Number of conflict events in Ghana (ACLED)");
ax

From 2010 to Present

Base on previews trend, let subset the data and analyze from 2010 to present day

Code

df=df[df["year"]>2009]

fig, ax = plt.subplots(figsize=(17, 7))

number_ev=df["event_type"].groupby([df['event_date'].dt.year.rename('year'), 
             df['event_date'].dt.month.rename('month')]).agg({'count'})

number_ev.index = [pd.datetime(year, month,1)
                   for (year,month) in number_ev.index]

number_ev.reset_index(inplace=True)
# plot
ax.plot('index','count', data=number_ev,color="b")

# Format the x axis 
ax.xaxis.set_major_locator(mdates.YearLocator(base=1))

ax.grid(True)
plt.ylabel("Number of conflict event")
plt.title("Number of conflict events in Ghana (ACLED)");
ax

Another view using Bar plot

Code

year_event = pd.crosstab(df.year,df.event_type).reset_index()

fig = px.bar(year_event, x="year", 
             y=["Protests","Riots","Violence against civilians","Battles","Strategic developments"],
             title=" <b> Political violence & protest events in Ghana (2010 - 13 nov 2021) <b>",
             labels={'value':'Number of Event','variable':'<b>Event Type'} ,
            # width=1000, # Width of map
            # height=600
            )
fig.update_layout(
    # barmode='group',  # comment this line to see stacked bar
    titlefont={'color':None, 'size': 15, 'family': 'San-Serif'},
    xaxis_tickangle=0,
    xaxis = dict(
        tickmode = 'linear',
        tick0 = 1,
        dtick = 1
    )
)

We can see that during 2019 and 2020 each event increase,specialy Protest, Riots,and Battles

Code

# event distribution by event_type and county

county_event_count=df.pivot_table(index="event_type",
                                  columns=["admin1"],
                                  values="event_date",
                                  aggfunc="count",)
# county_event_count

fig = px.imshow(county_event_count.fillna(0), 
                text_auto=True,
                # width=1000,
                # height=500,
               aspect="auto",
               title="Heatmap event distribution by event_type and county")
fig.update_layout(xaxis_tickangle=90,)

Protests event occurs most of the time in Greater Accra where Accra Ghana’s capital is and Ashanti
Most of Battles occurred in the North part of Ghana ,North East and Northern

Fatalities trend

Code

# fatalities trends using event_date

fig, ax = plt.subplots(figsize=(17, 7))

number_fatalities = df["fatalities"].groupby([df['event_date'].dt.year.rename('year'), 
             df['event_date'].dt.month.rename('month')]).agg({'sum'})

number_fatalities.index = [pd.datetime(year, month,1)
                   for (year,month) in number_fatalities.index]

number_fatalities.reset_index(inplace=True)

ax.plot('index','sum', data=number_fatalities,color="r")

# Format the x axis 
ax.xaxis.set_major_locator(mdates.YearLocator(base=1))

ax.grid(True)
plt.ylabel("Number of Reported Fatalities")
plt.title("Number of Reported Fatalities in Ghana (ACLED) ");

ax

Let have a better look with bar plot

Code

# fig, ax = plt.subplots(figsize=(17, 7))
# Total repoted fatalities by year
year_fatalities=df.groupby(['year',])[['fatalities']].sum()
year_fatalities.plot(kind="bar",figsize=(13,7),legend=False,color='dodgerblue')
a=list(year_fatalities.fatalities)
for index,i in enumerate(a):   
    y= str(round(i))
    plt.annotate(y,xy=(index,i),xycoords='data',bbox=dict(boxstyle="round", fc="white", ec="gray"),
             xytext=(0,-2),textcoords='offset points', ha='center')
sns.set_style("white")
plt.xlabel("Year ")
plt.ylabel("repoted fatalities ")
# plt.xticks(rotation=0)
plt.xticks([0,1,2,3,4,5,6,7,8,9,10,11],['2010','2011','2012','2013','2014',
     '2015','2016','2017','2018','2019','2020','2021 \n (1 Jan - 13  NOV)'] ,rotation=0);
plt.title("Total repoted fatalities by year", );
sns.despine()
plt.show()

Code

fig = (make_subplots(specs=[[{"secondary_y": True}]])

.add_trace(go.Scatter(y=number_fatalities["sum"],
                         x=number_fatalities["index"],
                         name = 'Fatalities',
                         line=dict(color='red', width=4, dash='dot')
                         )
             )
.add_trace(go.Scatter(y=number_ev["count"],
                         x=number_ev["index"] ,
                         name = 'Event count',
                         line=dict(color='darkblue', width=2, dash='solid')
                         ),
              secondary_y=True,
             )

.update_layout(title='<b> Fatalities & Event in Ghana<b>',
                  xaxis_title='<b>years<b>',
                  yaxis_title='<b>Reported fatalities <b>',
                  titlefont={'size': 15, 'family':'Serif'},
                  template='simple_white',
                  showlegend=True,
                  paper_bgcolor='#edeeee',
                  plot_bgcolor='#edeeee',
                #   width=1000,
                #    height=500,
                 )

)
fig.update_yaxes(title_text="<b> Event count</b>", secondary_y=True)

Interaction

Let zoom-in in year 2017 and see what happen

Code


# Total fatalities  by event_type and county

E2017=df[df["year"]==2017] .pivot_table(index="interaction",
                                  columns=["event_type"],
                                  values="fatalities",
                                  aggfunc="sum",)
E2017.drop([col for col, val in E2017.sum().iteritems() if val < 1], axis=1, inplace=True)
E2017.drop([col for col, val in E2017.sum(axis=1).iteritems() if val < 1], axis=0, inplace=True)
# E2017

fig = px.imshow(E2017.fillna(0), 
                text_auto=True,
                width=600,
                    height=600,
               aspect="auto",
              title="<b>Heatmap of total fatalities by event_type and interaction in 2017<b>")
fig.update_layout(xaxis_tickangle=45,
                 titlefont={'size': 15, 'family':'Serif'},)

We can see that during year 2017 Battles and Violence against civilians cause the most death. From the heatmap the interaction between communal militia against communal militia cause the most death in Battles and also the most death of civilians is perpetrator by communal militia. This show that , communal militia cause more than 85% of the total reported fatalities in 2017.

Let look at all the ten past years

Code

## Total fatalities  by event_type and county

event_interaction = df.pivot_table(index="event_type",
                                  columns=["interaction"],
                                  values="fatalities",
                                  aggfunc="sum",)
event_interaction.drop([col for col, val in event_interaction.sum().iteritems() if val < 1], axis=1, inplace=True)
event_interaction.drop([col for col, val in event_interaction.sum(axis=1).iteritems() if val < 1], axis=0, inplace=True)
# event_interaction

fig = px.imshow(event_interaction.fillna(0), 
                text_auto=True,
                # width=1000,
                #     height=500,
               aspect="auto",
              title="<b>Heatmap of total fatalities by event_type and interaction<b>")
fig.update_layout(xaxis_tickangle=45,
                 titlefont={'size': 15, 'family':'Serif'},)



# Total fatalities  by event_type and county

Code

county_fataliy=df.pivot_table(index="event_type",
                                  columns=["admin1"],
                                  values="fatalities",
                                  aggfunc="sum",)

fig = px.imshow(county_fataliy.fillna(0), 
                text_auto=True,
                # width=1000,
                #     height=500,
               aspect="auto",
              title="<b>Heatmap of total fatalities by event_type and county<b>")
fig.update_layout(xaxis_tickangle=90,
                 titlefont={'size': 15, 'family':'Serif'},)

Battles is the event with high reported fatalities during this past ten years, following by violence against civilians which occurs most of the time at the North part of Ghana, the center part and the capital as well

Let look deeper into Battles

Code

Battles = df[df["event_type"]=="Battles"]
battles_event = pd.crosstab(Battles['year'],Battles['interaction']).reset_index()
# battles_event

fig = px.bar(battles_event, x="year", 
             y=battles_event.columns[1:],
             title="<b>Interaction in Battles events in Ghana (2010 - 13 nov 2021)<b>",
             labels={'value':'Number of Event','variable':'<b>Interaction'} ,
            # width=1000, # Width of map
            # height=600
            )
fig.update_layout(titlefont={'size': 15, 'family':'Serif'},
    # barmode='group',  # comment this line to see stacked bar
    xaxis_tickangle=0,
    xaxis = dict(
        tickmode = 'linear',
        tick0 = 1,
        dtick = 1
    )
)

Most of battles each year are communal militia against communal militia (inter and intra communal violence), except in 2010 and 2011.
This show instability in the country at local level. The multitude of ethnic armed clash each year is not negligible when Ghana country, compared to many other African countries, has been described as a peaceful and stable country.

let look at the actor involved in those ethnic conflict

Code


ethnic_conflict=df[df["interaction"]=="communal militia versus communal militia"]

# let combine actor1 and actor2 together 
act = ethnic_conflict["actor1"].to_list() +  ethnic_conflict["actor2"].to_list()
act=pd.DataFrame(act,columns=["actor"])

fig, ax = plt.subplots(figsize=(13, 7))

# Set the color palette in reverse
colors = sns.color_palette('Reds', len(act.actor.value_counts().index[0:20]))
colors.reverse()



# plot Ethnic conflict to 20 actor involved
ax = sns.countplot(y=act.actor, order =act.actor.value_counts().index[0:20], palette=colors)
ax.set(ylabel=None, xlabel=None,);
plt.title("Ethnic conflicts - Top 20 actor involved");
sns.despine()
plt.tight_layout(pad=1, w_pad=1, h_pad=1)
ax

Top 20 actor involved in those ethnic conflict are minority ethnic groups in Ghana

View project repo

https://github.com/AbdelTID/Armed-Conflict-Event