Background
A data visualization was created to reveal the spatio-temporal patterns of armed conflict in selected South-east Asia countries between 2015 to 2020. The visualization is based on data provided by The Armed Conflict Location & Event Data Project (ACLED) at https://acleddata.com/#/dashboard.
The purpose of this post is to critique the current visualization and to propose an alternative visualization and interactive techniques to improve the current design.
Contents
A. Critique of Data Visualization Provided
B. Proposed Alternative Visualization and Interactive Techniques
C. Proposed Data Visualization
D. Preparation of Data Visualization
E. Key Observations from the visualization prepared
A. Critique of Data Visualization Provided
The original visualization created is shown below.
Figure 1: Original Data Visualization
Based on the original data visualization provided, the following are some comments on the clarity, aesthetics and interactivity aspects of the visualization.
Clarity
1. Spatial boundary of the selected country is not obvious.
While the title of the graph on the left specifies the country selected, the area occupied by the selected country is not evident on the map. A darker boundary or highlight of the region to make it more prominent will allow the reader to be better able to visualize the area occupied by the selected country.
2.Vertical axis label of the time series charts is not appropriate.
Currently the axis is labelled as “Count of Sheet 1” which is not an appropriate labelling for the vertical axes of the time series charts on the right. A more appropriate label could be “Number of Occurrences”.
3. Time series charts on the right did not indicate clearly which region or country the data represented is for.
There is no clear indication whether the time series charts on the right represented data of the selected country, the entire region (all Southeast Asian countries) or even the world.
4. Overview of Country Events needed.
Currently the visualization provides a detailed view of the locations of occurrence of political events of each country. An overview of the occurrences of political events of each country would allow for easy inter-country comparison.
5. No write-up to provide a context to the visualization.
There should have been a brief write-up to provide the user some context to the visualization and the key messages to take away from the visualization.
Aesthetics
1. Colours used to represent the events should be consistently applied across charts.
For the time series charts on the right, the colours used for the time graphs should be consistent with the colour scheme selected for the various event types for the geospatial chart on the left.
2. Labels of events could be of a stronger colour and bolded
The event types of the time series charts on the right should be labelled more prominently (either via a darker colour or bolded).
3. Selected country name peppered by dot plots.
For the country that is selected, dots representing events are plotted on the map on the left. However in doing so, the name of the country is covered over by the dots and the name of the country is not legible. Although the name of the country selected is still visible in the title, it would be aesthetically more pleasing to either have the name of the country removed or placed in such a way that it is not within the area of the selected country.
4. Colours of events to correspond to event categories as defined by ACLED
Colours selected for event types could correspond to different categories of event types with events within each category (i.e. Violent Events, Demonstration Events, Non-Violent Events) being of the same colour but in different hues. This will allow users to get an indication at first glance on the nature of the main types (categories) of events that have taken place in various countries.
Interactivity
1. Visualization allows user to select individual countries for viewing.
This is good as users will most likely want to view and compare events and occurrences of and between countries. Assuming that the chart on the right changes and corresponds to the country selected, the user is also able to visualize the developments of events over time for various countries. This feature should be included if not already so.
2. Visualization should allow user to select desired time periods to be displayed.
The visualization should provide interactivity for the user to select the time periods for which the data is displayed, for deeper analysis of changes in events over time. Both the spatial chart and line charts should respond to changes in selected time periods and display the relevant data.
3. Visualization could allow data of multiple selected countries to be displayed.
The visualization should be able to allow the user to view the data for multiple selected countries in the event that comparison between countries is desired. For the time series chart on the right, individual countries’ data should be a different line series.
4. Details of sub-regions to be displayed on mouse-over.
When user mouse-overs a particular sub-region of the selected country, details specific to the sub-region (e.g. sub-region name, number of various events taken place during the time period) should appear. This will provide greater understanding of how the occurrence of events are distributed geographically within the selected country.
B. Proposed Alternative Visualization and Interactive Techniques
Description of and Rationale for Proposed Visualization
The visualization is designed to have 2 dashboards, an Overview dashboard of countries involved in the data, and a Country Details dashboard to allow in depth view of events by country.
1. Overview Dashboard
Figure 2: Sketch of proposed Overview dashboard
1. Overview Dashboard to provide users with context of the data.
By providing users with an Overview Dashboard at the start, together with a short write-up, users will have a clearer understanding on what is the context of the data, and how to use the visualization to glean useful insights.
An Instructions for Use panel will also be provided for Users to be familiar on how to navigate between the dashboards.
2. Filter for Event Types to allow users an overview of the occurrences of various events in different countries
A filter to select different event types will be provided so that users are able to select event types of interest and have an overview of which event types have taken place more frequently in different countries as reflected by darker hues of red.
Red was selected as the main colour to represent the frequency of events taking place in the countries since the events mainly involve conflicts and sometimes violence.
For example, when user selects event types that are Demonstration Events (i.e. Protests and Riots), and if countries such as Indonesia will be shaded in darker hues of red, this means that Indonesia has greater numbers of Demonstration Events compared to the other countries. Conversely, if Violent Events (i.e. Battles, Explosions/Remote Violence, Violence against civilians) is selected, countries like Myanmmar or Cambodia will likely be coloured in darker hues of red compared to countries like Malaysia, indicating that there is a higher frequency of such events in countries like Myanmmar and Cambodia.
3. Frequency of occurrence of each Event Type in each country is shown when moused-over individual countries
The user will be able to have a view of the occurrences of events by event types in each country when he mouses-over the corresponding countries. A bar chart indicating the number of instances of event type will be provided in the mouse-over tooltip window.
4. Link to Country Details dashboard
To get more in depth view of the events that have taken place in each country, the user will click on the country in the map and will be brought to the Country Details dashboard to visualize the details by country selected.
2. Country Details Dashboard
Figure 3: Sketch of proposed Country Details dashboard
5. Geo-spatial plot of every event allows users to observe any geo-spatial patterns in events
Every event is plotted on the map on the left and colour coordinated with its event type. Users will be able to do a visual observation on whether there are any patterns in the geographical distribution of the events by event types for the selected country.
6. The line graphs on the right show the number of instances for each event type for the specified years
The line graphs on the right display the numbers of events by event types that had occurred year to year for the selected country.
7. Tool for viewing the data year by year will allow users to observe the changes and developments in events over time. By including a tool to allow users to playback or step through manually each year, while observing the corresponding points on the map and line charts, users may be able to spot temporal or spatial patterns in the data.
8. Inclusion of event details in the tooltips.
Tooltips will be created so that when users mouse-over each event on the map, the event details such as location of the event, number of fatalities and notes of the events are displayed.
C. Proposed Data Visualization
The proposed data visualization is found at the following URL:
https://public.tableau.com/profile/daniel.lin.yongyan#!/vizhome/DatavizMakeover3_16162605639950/OverviewDashboard?publish=yes
A screenshot of the dashboards are as follow.
Figure 4: Proposed Overview dashboard
Figure 5: Proposed Country Details dashboard
D. Preparation of Data Visualization
D1. Preparation of Dataset
D1.1 Selection of variables from ACLED dataset provided
The following variables were removed using EXCEL (right-click columns > delete) as they would not be used in the visualization – TIME_PRECISION, REGION and TIMESTAMP.
Figure 6: Removed TIME_PRECISION
Figure 7: Removed REGION
Figure 8: Removed TIMESTAMP
D1.2 Importing of dataset to Tableau
The above saved dataset was imported to Tableau by dragging from Windows Explorer to Tableau as shown below.
Figure 9: Import dataset to Tableau
D1.3 Changing variables to the correct data types
Some of the variables in the imported dataset were not classified as the correct datatypes. The following variables were re-classified accordingly as shown below.
- Year – classified as Date type
Figure 10: Reclassify YEAR to Date type
- Admin1, Admin2, Admin3 and Location – assigned with Geographic Role as State/Province.
Figure 11: Assign LOCATION with Geographic Role as State/Province
More importantly was to check that Latitude and Longitude variables had been assigned their correct geographic roles.
Figure 12: Check Latitude has been assigned Geographic Role > Latitude
Figure 13: Check Longitude has been assigned Geographic Role > Longitude
D2 Creation of Worksheets
D2.1 Creation of Overview Worksheet
D2.1.1 Pull Longitude, Latitude and Country onto workspace
Pull Longitude (generated) and Latitude (generated) into the Columns and Rows fields respectively, followed by Country into the Detail panel to create the regions of the countries in the dataset.
Figure 14: Bring Country, Latitude (generated) and Longitude (generated) into workspace
D2.1.2 Creation of filter to colour code countries by various event types
Pull Event Type to Colour panel.
Figure 15: Bring Event Type to Colour panel
Change Measure of Event Type to Count as shown below.
Figure 16: Change Measure of Event Type to Count
Now the countries are coloured based on the number of events (Count) that occurred in as shown below.
Figure 17: Countries coloured based on number of events
To allow users interactivity to view the number of events (by event types) that have occurred in the various countries, a filter for event type is created as follows.
Bring Event Type into the Filters panel.
Figure 18: Create filter for Event Type
Select all the event types in the dialog box that appears.
Figure 19: Select all event types
Select “Show Filter” for filter to appear on the right for users to select Event Types.
Figure 20: Select Show Filter
Use Edit Colours by clicking on the down arrow in CNT (Event Type) on the right to change the colour used. Red is selected since the data being reflected is the number of political conflicts, with greater intensity of redness being applied to countries with greater number of such conflicts.
Figure 21: Edit Colours
Edit the title of the Legend as shown below.
Figure 22: Edit Legend Title
Pull Country to Label panel to label the countries on the map.
Figure 23: Pull Country to Label panel
Now, users are able to have an overview on the occurrence of political conflicts in the various Southeast Asian countries by selected event types.
Figure 24: Overview by Demonstration Events
Figure 25: Overview by Violent Events
D2.1.3 Creation of Tooltip to provide more details for each country during mouse over.
On a separate worksheet (CtryEvents), pull Event Type into both Rows and Columns fields. For the Event Type in the Columns field, change Measure to Count to get the following bar chart.
Figure 26: Create bar charts for Event Types
As the events are categorized into Violent Events, Demonstration Events and Non-Violent Events, the order of the event types and colour coding should reflect the categorization. To do so, shift the Violence against civilians up after Explosions/Remote violence. Bring Event Type into the Colour panel and change the colours of the bars such that events in the same categories are of the same colour but different hues.
Figure 27: Colour bar charts by Event Types
Double click on the colour squares in the legend on the right to change colours of the bars.
Figure 28: Assign similar colours to event types of the same categories
Right click on Event Type to hide field label.
Figure 29: Hide Event Type label
Change horizontal axis label by double clicking on the “Count of Event Type”.
Figure 30: Change horizontal axis label
Double Click on the title of the chart to edit the title.
Figure 31: Update chart title
Add data labels to the bar chart by double clicking on the bars and under Mark Label, select “Always Show” as follows.
Figure 32: Update settings to always show data labels
On the Overview worksheet, click on the Tooltip panel to open up the Edit Tooltip dialog box. Insert the created bar chart as Tooltip as below and edit the Tooltip Dialog box as follows.
Figure 33: Insert the created bar charts as part of Tooltip on the Overview sheet
Figure 34: After inserting the worksheet into Tooltip
The following is the resulting display when the mouse is brought over each country.
Figure 35: Tooltip provides country info when moused-over
As the Tooltip window is too small to display the chart, resizing is needed on the CtryEvents sheet to ensure the chart fits into the Tooltip window.
Figure 36: Re-size bar chart to ensure fit in Tooltip window
The dimension of the Tooltip window is also edited to allow a larger Tooltip to be displayed.
Figure 37: Edit size of bar chart to fit in Tooltip window
Update the title of the Overview by double-clicking on the title.
Figure 38: Updated title of Overview worksheet
D2.1.4 Insertion of filter for Year
Pull Year into the Filters pane as shown and select Year in the resulting Filter Field window that appears.
Figure 39: Create Year filter
Figure 40: Select Years to be displayed
In the following window, select the relevant years that we are interested in – 2015 to 2020.
Figure 41: Select Years 2015 to 2020
Update the title of the graph with the period of interest (2015-2020) and the resulting appearance of the dashboard is as follow.
Figure 42: Resulting display
D2.2 Creation of Detailed Worksheet
D2.2.1 Bring Event Id Cnty, Latitude and Longitude onto the workspace.
Pull Event Id Cnty onto the Detail panel, Latitude and Longitude onto the Rows and Columns field as shown below.
Figure 43: Pull Event Id Cnty, Latitude and Longitude in workspace
Pull Event Type onto the Colour panel to get the following screen.
Figure 44: Colour by Event Type
D2.2.2 Create filter for Event Type to allow users to view events by event types.
Pull Event Type into the Filters pane as shown below.
Figure 45: Create Event Type filter
Select all in the Filter window that appears.
Figure 46: Create Event Type filter
Select “Show Filters” at the Event Type in the Filters panel.
Figure 47: Show Event Type filter
D2.2.3 Create a filter for Country to allow users to view events by country.
Pull Country to the Filters panel.
Figure 48: Create Country filter
Select all the countries.
Figure 49: Select all countries
Select Show Filter in the Filters panel for Country.
Figure 50: Show Country filter
Now, if a country is selected, the events for the specified country appears. However the boundary of the country is not obvious as shown.
Figure 51: Show Country filter
D2.2.4 Create a layer for Country to highlight the region and outline of the country selected.
Pull the Country dimension onto the workspace to “Add a Marks Layer”.
Figure 52: Pull Country onto workspace
The result is a point representing the Country location.
Figure 53: Blue dot represents Country location
In the Country panel, change Marks type from Automatic to Map as shown.
Figure 54: Change Marks type to Map
The following is the resulting display.
Figure 55: Shaded region represents Country area
As the points need to appear above the highlighted area, pull the Country pane to be below the Latitude pane. The following is the resulting display.
Figure 56: Resulting display where country selected is highlighted
Pull the Country variable into the Detail panel and remove the colour by Country icon as shown so that the countries are highlighted with the same colour.
Figure 57: Pull Country to details panel
Figure 58: Remove colour by Country
Click on the Colour panel to edit the colour of the highlight.
Figure 59: Edit colour of Country highlight
The resulting display is as shown below, where the region of the selected country is highlighted in pink with events plotted on top. The boundary of the selected country is also clearly demarcated.
Figure 60: Resulting display
D2.2.5 Creation of an interactive slider for users to scroll through each year
By adding an interactive tool to scroll through individual years, the user will be able to view the development and changes of the occurrence of events.
This is done by pulling Year to Pages.
Figure 61: Pull Year to Pages
A tool on the right appears, allowing users to select the year that they are interested in and to also play a slideshow that will display the events in a chronological manner.
Figure 62: Use tool on the right to select year or to play slideshow
D2.2.6 Make circles representing events more distinct.
Currently the circles plotted for each event is overlapping and does not appear in a clear manner. To make the circles more distinct, first select Circle in the Marks drop down list as shown.
Figure 63: Select Circle in the Marks drop down list
Then click on the Colour panel and then on the Border colour to select a dark colour for the border of the circles representing the events as shown.
Figure 64: Select Circle colour and border
E. Key Observations from the visualization prepared
1. Philippines has the most occurrences of political events while Laos has the least.
Philippines has the greatest number of political events while Laos has the lowest number of political events as seen in the following screenshot below where Philippines has the darkest hue of red and Laos has the lightest hue.
Figure 65: Colour intensity reflecting number of political events
2. Philippines has the greatest number of Violent events, followed by Myanmar and Thailand.
Based on the Overview dashboard, Philippines has the greatest number of violent events (i.e. Battles, Explosions / Remote violence, and Violence against civilians), followed by Myanmar and Thailand. This is observed from the colour intensities of the mentioned countries in comparison to their neighbours when the Violent Events have been selected in the Event Type filter.
Figure 66: Philippines has most number of Violent Events
Interestingly, the most frequent type of Violent Event is different for the 3 mentioned countries. For Philippines, the most frequent type of Violent Event is Violence against civilians as seen below.
Figure 67: Main violent event type is Violence against civilians
For Myanmar, the most frequent type of Violent Event are Battles.
Figure 68: Main violent event type is Battles
Whereas for Thailand, the most frequent Violent Event type is Explosions / Remote Violence.
Figure 69: Main violent event type is Explosions / Remote Violence
3. In terms of Demonstration Events, Indonesia had the most followed by Myanmar, while Laos had the least. Protests made up the main component of Demonstration Events.
From the Overview dashboard, when Demonstration Events (Protests and Riots) were selected, it is observed that Indonesia had the greatest number of such events since it had the darkest hue of red, followed by Myanmar. Laos once again had the lightest hue and hence the lowest number of Demonstration Events.
Figure 70: Indonesia had most number of Demonstration Events
When moused-over Indonesia, it is observed that the majority of Demonstration Events is made up of Protests with number of Riots being about one fifth as much as number of Protests.
Figure 71: Majority of Demonstration events are Protests
The same trend is observed in the other countries, which can be observed when we mouse-over the other countries.
Figure 72: Majority of Demonstration events are Protests
4. In the Philippines, the number of Violent Events decreased over time, mainly led by the fall in incidents involving Violence against civilians and mainly in the Northern region.
From the Country Details dashboard, we are able to view the drop in Violent Events over time was mainly due to the fall in Violence against civilians as shown.
Figure 73: Number of Violent Events fell over time
From the dashboard, when we step through year by year from 2016 to 2020, it is observed that there is an easing up on violent events in the North, more so than the South as shown below.
Figure 74: Number of Violent Events fell over time
5. Number of Demonstration Events in Indonesia fell initially before climbing significantly from 2017.
The number of Demonstration Events (i.e. Protests and Riots) in Indonesia had a falling trend from 2015 to 2017 but experienced a sharp increase from 2017 to 2020 and this increase is led mainly by the large increase in the numbers of Protests as shown below.
Figure 75: Number of Demonstration Events fell before rising sharply
In terms of geographical patterns of the Demonstration Events, there was a slight shift over time in the concentration of incidents from the Southern Island of Java, outwards towards the Western island of Sumatra, Northern Island of Sulawesi and Western New Guinea in the East as shown below.
Figure 76: Outward shift in concentration of Demonstration Events