by Bhaskar Ghosh, Shuai Hao, Kishan Sheth and Hussain Vahanvaty
For the final project of our Data Visualization class (Spring ’19) we wanted to focus on issues which have a global context and a dynamic nature. The terrorist attacks in New Zealand in March 2019 prompted us to look at terrorism-related data and how it has evolved over the years.
Terrorism has long been one of the biggest threats to safety and stability across the world. Different forms of terrorism exist, including state-sponsored, political and religious terrorism. Methods and tactics of terror groups have evolved over the years. To understand in what ways the transformation has taken place, we looked at the Global Terrorism Database (GTD).
- Static visualizations and stories about terrorist attacks in Europe by Washington Post
- Heat map visualization about global terrorist attacks year by year by Business Insider
- The article on ‘Global Terror and the EDA Visualization Rabbit Hole’ used Tableau to make simple visualizations like line charts, bar charts, scatter plots.
- ‘Global Terrorist Attacks’ used Tableau to visualize the same database by creating an interactive map showing all terrorist attacks around the world.
Our analysis was based on the Global Terrorism Database (GTD). One of the most comprehensive public databases on terrorism, it contains data about terrorist attacks from 1970 to 2017. This is how GTD has defined terrorism: “the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation.”
Link to GTD’s documentation: https://www.start.umd.edu/gtd/downloads/Codebook.pdf
We used the following fields from the data:
- Year (iyear)
- Month (imonth)
- Day (iday)
- Country (country, country_txt)
- Region (region, region_txt)
- Weapon Type (weaptype1, weaptype1_txt)
- Target/Victim Type (targtype1, targtype1_txt)
- Total number of fatalities (nkill)
- Total number of injured (nwound)
- Perpetrator Group Name (gname)
Analysis: We wanted to focus on the number of attacks in different regions across the world, so we decided to use a map-based visualization for that purpose. We chose to divide the time period between 1970–2017 into 5 decades. We generated the GeoJSON for the map visualization and then used Pandas (Python) to create another JSON structure which had the number of fatalities per country per decade from 1970–2017. We fused the two JSON structures together to create one structure for our visualization.
Using GTD, we created another JSON for showing the deadliest attacks, which had the following structure:
[ 1, “2001–09–11”, 2996.0, “Al-Qaida”],[ 1, “1985–06–23”, 329.0, “Sikh Extremists”]
Entries in the JSON above have the following order: decade, date, fatalities, group name
We analyzed the GTD to determine hard and soft targets and how target types have evolved over the years. We determined that hard targets like government officials and military personnel have become harder to attack. Private citizens have bore the brunt of most of terrorist attacks, especially since 2011.
We also decided to look at the evolution of different weapon types. Using Pandas, we isolated the weapon types that are most in use by terror groups. Unsurprisingly, we found that firearms and explosives are the weapons of choice and have been used to carry out most of the deadliest attacks.
We used the following formula to determine the lethality of weapon types:
Lethality of weapon (for a given year) = (no. of people killed + no. of people wounded) / no. of attacks in a year
Techniques: We used Pandas, Matplotlib and Tableau for most of our data analysis. We started out by making rough visualizations on Tableau. It helped us to spot trends in the data and to decide the visualizations we should use for the final project.
Algorithms: We did not use any specialized algorithms for our analysis. Most of our analysis was done using standard statistical techniques for drawing line charts and linear regression.
For the map visualization, we have the following visual encoding:
For the bubble chart, we have the following visual encoding:
For the first line chart, we have the following visual encoding:
For other line charts, we have the following visual encoding:
As for taxonomy of interactions, we only used data manipulation. We used point selection on our world map and bubble chart, showed animations when generating charts and highlighted different data as users scroll.
Alternatives we abandoned:
We tried to use dots to represent each attack at first, but the total number of attacks is large and we are not able to write the narration part with the changes of dots due to large amount of missing data.
Our final visual and interaction encoding are successful because most audience praised our project. A few audience feel confused the title of our visualizations should change as users scroll. Our visualization gives the audience deeper understanding about the evolution of terrorism.
- Since we didn’t fully understand how Scrollama library works, more than one chart gets appended to the same SVG when users scroll too fast.
- Due to time constraints, we didn’t make the website responsive, so users won’t have a good experience if they access the visualization on mobile devices.
- The GTD is huge database with a lot of valuable information for us to dig into. We would extend our project by analyzing more parameters like terrorist groups and attack types
Acknowledgements or References
We used the following websites for the text of our visualizations:
Here’s the link to our final visualization: