1. Overview
This post will show how to visualise the spatio-temporal patterns of armed conflict in selected South-east Asia countries between 2015-2020. The original chart can be found at South-east Asia Armed Conflict Analysis Firstly, we will criticize the original charts then show how to make better visualization from the same data set.
1.1. Original chart
In this Analysis, below codebook and data files were used for analysis.
- Raw Data Southeast-Asia_2010-2020_Oct31.xlsx
- Armed Conflict Location & Event Data Project (ACLED) Codebook codebook
1.2. Skim of raw data
| Name | raw_df |
| Number of rows | 38244 |
| Number of columns | 29 |
| _______________________ | |
| Column type frequency: | |
| character | 16 |
| numeric | 12 |
| POSIXct | 1 |
| ________________________ | |
| Group variables | None |
Variable type: character
| skim_variable | n_missing | complete_rate | min | max | empty | n_unique | whitespace |
|---|---|---|---|---|---|---|---|
| EVENT_ID_CNTY | 0 | 1.00 | 4 | 8 | 0 | 38244 | 0 |
| EVENT_TYPE | 0 | 1.00 | 5 | 26 | 0 | 6 | 0 |
| SUB_EVENT_TYPE | 0 | 1.00 | 5 | 35 | 0 | 25 | 0 |
| ACTOR1 | 0 | 1.00 | 7 | 107 | 0 | 314 | 0 |
| ASSOC_ACTOR_1 | 28700 | 0.25 | 7 | 1213 | 0 | 1336 | 0 |
| ACTOR2 | 12558 | 0.67 | 7 | 97 | 0 | 274 | 0 |
| ASSOC_ACTOR_2 | 27948 | 0.27 | 6 | 390 | 0 | 998 | 0 |
| REGION | 0 | 1.00 | 14 | 14 | 0 | 1 | 0 |
| COUNTRY | 0 | 1.00 | 4 | 11 | 0 | 8 | 0 |
| ADMIN1 | 0 | 1.00 | 3 | 47 | 0 | 252 | 0 |
| ADMIN2 | 20 | 1.00 | 3 | 26 | 0 | 1295 | 0 |
| ADMIN3 | 2917 | 0.92 | 2 | 28 | 0 | 3370 | 0 |
| LOCATION | 0 | 1.00 | 2 | 30 | 0 | 6573 | 0 |
| SOURCE | 0 | 1.00 | 2 | 168 | 0 | 2166 | 0 |
| SOURCE_SCALE | 0 | 1.00 | 5 | 25 | 0 | 20 | 0 |
| NOTES | 0 | 1.00 | 30 | 1609 | 0 | 36441 | 0 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| ISO | 0 | 1 | 4.445500e+02 | 264.34 | 104.00 | 1.040000e+02 | 6.080000e+02 | 6.080000e+02 | 7.640000e+02 | ▇▃▁▇▆ |
| EVENT_ID_NO_CNTY | 0 | 1 | 4.494230e+03 | 3117.74 | 1.00 | 1.734000e+03 | 4.091500e+03 | 7.027000e+03 | 1.127400e+04 | ▇▆▅▅▃ |
| YEAR | 0 | 1 | 2.016310e+03 | 2.91 | 2010.00 | 2.014000e+03 | 2.017000e+03 | 2.019000e+03 | 2.020000e+03 | ▃▃▆▇▇ |
| TIME_PRECISION | 0 | 1 | 1.060000e+00 | 0.28 | 1.00 | 1.000000e+00 | 1.000000e+00 | 1.000000e+00 | 3.000000e+00 | ▇▁▁▁▁ |
| INTER1 | 0 | 1 | 3.200000e+00 | 2.06 | 1.00 | 1.000000e+00 | 2.000000e+00 | 6.000000e+00 | 8.000000e+00 | ▇▂▁▅▁ |
| INTER2 | 0 | 1 | 2.920000e+00 | 3.07 | 0.00 | 0.000000e+00 | 1.000000e+00 | 7.000000e+00 | 8.000000e+00 | ▇▂▁▁▅ |
| INTERACTION | 0 | 1 | 3.175000e+01 | 19.96 | 10.00 | 1.200000e+01 | 2.700000e+01 | 6.000000e+01 | 8.000000e+01 | ▇▃▁▅▁ |
| LATITUDE | 0 | 1 | 1.174000e+01 | 8.64 | -10.17 | 6.570000e+00 | 1.336000e+01 | 1.794000e+01 | 2.773000e+01 | ▂▁▇▇▅ |
| LONGITUDE | 0 | 1 | 1.080200e+02 | 11.73 | 92.20 | 9.796000e+01 | 1.018600e+02 | 1.209800e+02 | 1.408600e+02 | ▇▂▃▂▁ |
| GEO_PRECISION | 0 | 1 | 1.490000e+00 | 0.56 | 1.00 | 1.000000e+00 | 1.000000e+00 | 2.000000e+00 | 3.000000e+00 | ▇▁▆▁▁ |
| FATALITIES | 0 | 1 | 6.700000e-01 | 2.97 | 0.00 | 0.000000e+00 | 0.000000e+00 | 1.000000e+00 | 2.430000e+02 | ▇▁▁▁▁ |
| TIMESTAMP | 0 | 1 | 1.578913e+09 | 14870918.64 | 1552576414.00 | 1.567539e+09 | 1.579013e+09 | 1.591125e+09 | 1.604424e+09 | ▅▃▃▇▂ |
Variable type: POSIXct
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| EVENT_DATE | 0 | 1 | 2010-01-01 | 2020-10-31 | 2017-04-13 | 3911 |
1.3. Data columns of raw data
1.4. Final Dashboard
Final Dashboard (Tableau link)
2. Critique of the original charts
2.1. Clarity
- According to classification of ACLED event types in the codebook, there are three “General” categories and each category can be divided into several “Event type”. However, the original map only shows “Event Type”, without showing “General” groups. For this reason, readers cannot easily recognise the difference, in terms of severity of events.
- Mark as a single dot: Every event is represented by single dot with different colors. However, if several events occured at the same place, it would be hard to count how many events were accured just by the map.
- Time: Also, single dots do not tell when the events occured. This map only shows the events occured between 2015 to 2020.
- “Count of Sheet1”: As seen below, “Count of Sheet1”, y-axis label is ambiguous, because readers do not know “Sheet1”. This label should be changed with proper name.
- “Event Date”: As seen below, the label of x-axis, “Event Date”, does not properly represent meaning of values of x-axis. The x-axis value only represents the year and there is no information about the sepecific date.
- Title of line graph “Armed Conflict Event by Type”: This title does not represent the whole meaning of the chart because it also shows the trend by time.
2.2. Aesthetics
- Location of legend: The legend locats on the right side, so it’s hard to understand at a glance while mapping each of dot and the legend.
- Range of y axis: As seen below graph, the ranges of each event’s y-axis are not consistant. Readers will be confused when looking at the graph because there the numbers are not uniform. The values should be compared up and down each of the graph.
- Color of line chart: As seen above, all the line graphs have the same color regardless of the event type.
- Country name on the map: When readers select a country, country’s name is disappear due to many dots. Country names should be recognized more clearly on the map.
2.3. Interactive
- Tool tip for a dot: When readers hover over to the dots, only four information appeared in a form of tool tip. However, more valueable information should be popped up for readers. Especially, latitude and longitude are not very helful for readers.
- Tool tip for a line graph: When readers hover over to the line graph, only three information appeared in a form of tool tip. As described above, more valueable information should be popped up for readers. Especially, “Event Type” and “Year of Event Date” are redundant because these values can easily find in y-axis and x-axis in a graph.
- Radio button for selection of country: In order to compare data by countries, legend(filter) should be changed from radio button to check box.
3. Alternative suggestion of the graph
3.1. Sktechs of graphical presentation and advantages
1) Sketch1 - Dashboard
In order to show the simliar level of “Event Type”, we can paint in similar colors within the same “General type”: “Violent events”, “Demonstrations”, and “Non-violent actions”.
In order to compare not only the number of events, but also the number of fatalities, we can create stack bar charts by countries and years. By doing so, we can easily recognise the changes in time at a glance.
We can make filters of “Year” and “Event Type” to compare values easily.
The range of x-axis is uniform, so readers also can easily compare the values at a glance.
Show differences in the number of events, we can place different sizes of dots according to the number of events occured on the locations.
Also, put filters of “Year”, and “Event type” on the right handside to see the specific information.
2) Sketch2 - Tooltips
- Bar charts’ tooltips: In order to show the trend of the number of events occured and fatalities, put a line graph in a tooltip function.
- Map’s tooltip: In order to show every elements in each location information clearly, put related fields and values into toolbox.
3.2. Final dashboard of visualisation using Tableau.
All the charts we designed above were rearranged for one cleaned dashboard.
Go to the Final Dashboard -> Final Dashboard
4. Step-by-step preparation of visualization
4.1. Upload raw data set in Tableau
- Upload raw data “Southeast-Asia_2010-2020_Oct31.xlsx” into Tableau “Data Source”.
4.2. Make necessary fields clean and transform for analysis
- Move up “Latitude” and “Longitude” filed from Measure pannel to Attribute pannel.
- Create a Hierachy “Region”, “Country” and “Adminn1” field for geospatial analysis.
- Make sure to change ’Admin1" field type to “Geographic Role” > “State/Province”.
- Create a group for aggregate “Event Type” for further analysis".
.png)
4.3. Create Map chart to show geospatial analysis
- Put all necessary fields into Pannel as below.
- Put “Country”, “Event Tpye”, and “Year”. Then click “Show Filter” to see in the filter pannel
- Go to legend area and change title of “Event Type” legend
- Go to “Year” legend and click small triangle right corner, then click “Single Value (slider)”
- Go to Tool tip menu in “Marks” pannel and edti as below
4.4. Create bar chart to show trend of number of events comapring by countries
- Put all necessary fields into Pannel as below.
- Click “Show Filter” to show in the filter pannel.
- Right click x-axis and click “Edit Axis” then delete “Title”.
- Right click label of columns, “Event Date”, and check “Hide Field Labels for Columns”
- Go to filter and change the legend title
- Right click title, “Sheet 2(2)” and check “Hide Title”
- Sort “Country” by “CNT(sheet1)” field
- Go to filter pannel and click small triangle in “Year” filter. Then check “Multiple Values (dropdown)”
- Change order of “Event Type” by clicking “Sort” > “Manual” option.
4.5. Create bar chart to show trend of number of fatalities comapring by countries
- Duplicate “No of Event” sheet we created above, and rename “No of Fatality”
- Drag and drop “Fatalities” measures onto “CNT(Sheet1)” to change the target value.
- Again go to x-axis and right click. Then delete “Title”.
4.6. Create tooltips for trend of number of events and fatalities
- Duplicate “No of Event” sheet we created above, and rename “No of Event_tooltip”
- Click “+” sign and extend fileds to “Month”
- Drag and drop all fields as below Then change graph type to “Line”
- Drag and drop “Event Type” and “CNT(Sheet(1))” onto “Label” in Marks. Click “Label” button and change setting as below.
- Duplicate above sheet and rename to “No of Fatality_tooltip”. Then change “CNT(Sheet1(1))” to “Fatality” in Rows.
- Go back to “No of Event” sheet and click “Tooltip”. Then type as below.
- Go back to “No of Fatality” sheet and click “Tooltip”. Then type as below.
4.8. Create Dashboard
- Put all necessary fields into Pannel and readjust size of each chart as below.
- Click “More option” in “Year” filter and go to “Selected Worksheets…”
- Select every worksheet as below.
4. Do the same action as above for “Event Type” filter in “1. The number of armed conflict events and fatalities (Jan 2010 to Oct 2020)”
- Other filters in “2. Location and the number of armed conflict events (Jan 2010 to Oct 2020)” do not need to change filter setting.
5. Major observations
5.1. The number of events and fatalities by country and year
We can observe that the number of armed conflict events and fatalities changed over time. In Philippines, especially, we can see the trends decreasing over time in “Violence against civilians”. The number of fatalities was also dcreased between 2016 and 2020. However, “Protests” in Indonesia show an increasing trend from 2016 to 2020.
5.2. Frequency of the events occured in the same place (average)
Armed conflict events have taken place in various places in each country. we want see how often events occured in one location. As seen the table below, the value can be calculated as “The number of locations events event occured” divided by “The number of locations events event occured(Distinct)”. By using the value, we can observe that “Protests” occured more frequently in the same place than other events in Philippines and Indonesia. However, in Myanmar, “Battles” occured more frequently in the same place than other events.
