Introduction

Terrorism is a global challenge that impacts various regions and continues to change over time, making it important for policymakers and security agencies to understand its evolving trends. This report utilizes historical terrorism data to focus on analyzing the success rates of terrorist groups over time, assess casualties across regions, and explore the relationship of attack types with many variables. This dataset is interesting for exploring trends in global terrorism and identifying regions or countries with the most recorded terrorist attacks and casualties. It will also be interesting to observe trends between attack types, target types, and more. This dataset will allow graphs such as heatmaps for geographical trends, facetted plots to compare terrorist attacks across different countries and regions, bar charts to compare target type frequency (identifying patterns in target attacks based on geography or period), and more. Through these insights, this report aims to understand terrorist groups’ patterns and behavior better to counteract their attacks more effectively.

Data Description

The Global Terrorism Database (GTD) is a comprehensive dataset containing over 180,000 records of terrorism spanning the years 1970 to 2017, both domestically and internationally. Each row in the dataset represents a single terrorist incident, while the columns represent various aspects of the event. The dataset includes over 135 variables, with both categorical (e.g., terrorist group name, attack type, region) and quantitative data (e.g., number of casualties, attack success, year of event). Key variables in this dataset inform us of the data and location of the attack, methods, and targets of the attack, outcomes such as casualties and property damage, and more. The data consists of over 200 observations and adheres to the complexity requirements for producing diverse and informative graphs. The variables that will be focused on throughout the report are success, region, summary, nkills, region_txt, attacktype1_txt, gname, and country_txt.

Research Question

The four main research questions that we are trying to asnwered are the following: 1. What are the key themes and variables to consider when analyzing patterns and trends in this dataset? 2. What are the trends in the number of casualties across different regions over time, and which attack types caused the most damage? 3. How has the percentage of successful terrorist attacks changed for the most active terrorist groups globally? 4. Are certain attack types more likely to succeed in specific regions?

Graphs and Analysis

Question 1: What are the key themes and variables to consider when analyzing patterns and trends in this dataset?

When first analyzing this dataset, we wanted to identify prominent themes, words, or variables that could guide our focus. To do this, we created a word cloud to highlight such terms and reviewed the outputs using the summary of the dataset. One of the first themes we wanted to focus on was “casualties”, so we created a histogram to investigate the distribution of casualties by analyzing the variable “nkill”, which records the number of fatalities for each terrorism record.

The first graph we created is the word cloud, which provided us with valuable areas of interest. Words like “attack”, “casualties”, and specific terms such as “explosion”, “bomb”, and “fire” stood out to us. This drew our attention to “attack types”. Moreover, terms like “injury” emphasized the importance of focusing on “casualties” as well. Another set of words was country names, which piqued our interest in analyzing geographical distributions of terrorist incidents. Thus, this included looking into variables such as the groups involved and the regions they were based in. In conclusion, this exploratory step shifted our focus to topics such as attack types, casualties, and regional impacts.

The second graph we created is a histogram showcasing events with over 100 fatalities. This helps us observe the distribution of significant deaths caused by terrorist attacks while eliminating the skew caused by incidents with no casualties. We can observe that the majority of incidents resulted in around 100-300 casualties, with the largest peak being around 100 casualties. However, the data also highlights several outlier incidents where death counts reached into the thousands.

This histogram revealed the variability in casualty counts. While the large majority of attacks resulted in fewer than 300 deaths, a small number of highly deadly attacks exist. These few but devastating records showcase the dramatic impact a few successful terrorist attacks can have.

Question 2: What are the trends in the number of casualties across different regions over time, and which attack types caused the most damage?

We explored potential trends in the number of casualties from 1970 to 2017 and across and different regions over time and which attack types caused the most damage. To answer this, we examined the variables iyear (year), region_txt (region), and nkill (casualties per attack) to identify casualty trends. Additionally, we calculated total_kills, representing the total casualties aggregated by year and region, to provide a clearer view of regional trends and trends over time.

The above graph shows the number of deaths caused by terrorist incidents across various regions over time with each line representing a region. In particular, the number of terrorism-based deaths increases significantly for areas in the Middle East & North Africa, Sub-Saharan Africa, and South Asia after the 2000s. This trend aligns with the escalation of conflicts in these regions during this period. On the other hand, regions like Western Europe and North America maintained relatively low casualty counts over time. On the other hand, Central America & Caribbean showed a sharp increase in the number of casualties in the 1980s, but have since declined.

This graph might be useful when we want to focus on specific questions that will dive deeper into the regions with significant changes in trends or look into the reasons for the increase in deaths. It also highlights that terrorism impacts regions unequally, where regions that are less stable and have more conflicts often have more casualties. This emphasizes how important is it to allocate more resources to regions that are more impacted by terrorism.

To gain more insights into which attack types caused the most casualties across regions, we used the same variables from the previous graph: region_txt (region) and calculated total_kills and included a new variable attacktype1_txt (attack type) to analyze the impact of different attack strategies. Attack types include armed assault, assassination, bombing/explosion, facility/infrastructure attack, and more (see the x-axis of the graph below for all of the attack types).

The heatmap shows the total number of casualties (nkill) from various attack types such as armed assault, bombing/explosion, and assassination across world regions. The color gradient shows the severity of casualties in each region for each attack type, with green being low causalities, gray being medium, and red being high. The heatmap shows that bombing/explosion in the Middle East & North Africa has the highest casualties as it is the only box that is red on the heatmap, while other regions and attack types show lower impacts. Furthermore, the Middle East & North Africa have the highest number of boxes that are not green, which means that they incur a lot of casualties from various types of attacks. Additionally, armed assault seems to result in moderate casualties for around 50% of the regions, as indicated by the gray areas. This visualization is useful as you can quickly identify hotspots and regional trends.

Therefore, from 1970 to 2017, regions like the Middle East & North Africa, Sub-Saharan Africa, and South Asia saw significant increases in casualties, especially after the 2000s. On the other hand, regions like Western Europe and North America experienced stable and low casualties. Bombing/explosions caused the abnormally highest number of casualties in the Middle East & North Africa, which is an opportunity for counterterrorism groups to prioritize.

Question 3: How has the percentage of successful terrorist attacks changed for the most active terrorist groups globally?

We wanted to explore whether the percentage of successful terrorist attacks changed for the most active terrorist groups globally. Understanding the success rates of terrorist attacks provides insight into the operational effectiveness and adaptability of terrorist groups over time. This way, we can understand if certain groups consistently achieve their objectives, or if counterterrorism strategies, changes in organizational dynamics, and more affect these rates. To do this, we calculated the success rates of each terrorist group for each year. We grouped data by “gname” (terrorist group name) and “iyear” (year of attack) and calculated the success rate as the ratio of successful attacks to the total number of attacks within that year. Additionally, to analyze geographical regions impacted by groups, we used “country_txt” (country name) to identify where attacks occurred and mapped them to their respective regions on the world map.

The first graph uses a facet layout to showcase the success rates of the top 10 most active terrorist groups from 1970 to 2017. Each panel represents a single terrorist group, revealing trends in their operational effectiveness over the years.

From the graph, we can identify that groups like the New People’s Army (NPA) and Shining Path (SL) demonstrate high consistency in their success rates, often exceeding 90%. These groups also show the longest operational histories. Other groups, such as Al-Shabaab, Boko Haram, FMLN, ISIL, and the Taliban, have shorter operational periods and more volatile success rates. These groups tend to show a downward trend or have ceased activity. Groups like the PKK, IRA, and FARC exhibit more variability in success rates. For instance, periods of extremely high success rates are often followed by sharp declines, likely due to changes in counterterrorism efforts or internal organizational shifts.

This graph highlights stark differences in operational stability and effectiveness among the most active terrorist groups. It provides a basis for further analysis of historical, political, and military influences shaping these trends.

The second graph is a world map illustrating the geographical activity of the top five most active terrorist groups. By mapping the concentration of attacks, this visualization provides the spatial context of each group’s regional focus and helps us identify patterns of the attacks.

The top 5 terrorist groups are Al-Shabaab, FMLN, ISIL, Shining Path, and the Taliban. Al-Shabaab heavily concentrates on attacking East Africa, particularly Somalia, Kenya, and Ethiopia. This group has a relatively shorter operational period and fluctuating success rates, emerging likely due to its intense regional conflicts and counterterrorism measures such as AMISOM. The FMLN primarily attacks Central America, reflecting its localized political focus in El Salvador during its active years. We learned from the previous graph that the FMLN had a shorter and slightly volatile operational period, so its concentrated activity is justified. The group died down after the Salvadoran Civil War ended. ISIL activity is spread across Eastern Europe, Australasia & Oceania, the Middle East & North Africa, and Western Europe. The ISIL also had a shorter operational period with higher fluctuations in success, perhaps affected by how spread out their attacks were. Thus, resources were spread thin and its rise and fall was quick. The Shining Path dominated activity in South and Central America. This group had particularly high consistency in their success rates due to concentrating on Peru during its internal conflict with strong ideological alignments. The Taliban operates mostly in South Asia, particularly in Afghanistan and Pakistan. They also had shorter operational periods and more fluctuations in their success tied to the socio-political dynamics of Afghanistan. The involvement of the US military led to its subsequent withdrawal.

This geographical visualization showcases how local political dynamics, cultural contexts, and other factors could influence group effectiveness and longevity. Localized groups like the Shining Path and FMLN succeeded due to their localization and specific political conditions but struggled to maintain effectiveness after the campaigns ended. More dispersed groups such as the ISIL faced challenges in maintaining consistency. Groups like the Taliban and Al-Shabaab sustained themselves on intensive regional conflicts but quickly were subdued by counter-measures.

To answer the research question, the analysis reveals that the majority of the most active terrorist groups exhibit a general decline in success rates over time. This trend is mostly attributed to groups becoming inactive as many of the conflicts that drove their formation have been resolved or diminished in significance. Another factor that impacts this decreasing rate is the effectiveness of counter-terrorism measures in curbing these terrorist operations. Moreover, the challenges of operational resources being spread too thin across multiple country targets lead to lower success rates. Overall, the decreasing trend can be influenced by a variety of factors such as changes in political and military contexts, resource constraints, geographical concentration, and external pressures.

Question 4: Are certain attack types more likely to succeed in specific regions?

We wanted to learn about the relationship between attack types and their likelihood of success in different regions. The success variable in the dataset is binary (either 0 or 1). The two graphs that are used to answer this question will use the variables success, attacktype1_txt (attack types), and region_txt (name of each region), and region. Additionally, we will also perform a statistical analysis to see if there is a relationship between region and success.

First, we will start with a stacked bar plot shown above that shows the distribution of attack types across different regions, with the y-axis being the count of attacks and the x-axis showing the different attack types. The legend shows a different color for each region, which is reflected in the stacked bar plot.

The graph shows that the most frequently used attack type globally is bombing/explosion with over 75,000 recorded incidents. A significant portion of these attacks occurred in the Middle East & North Africa and South Asia, which might reflect the ongoing conflicts and instability in these regions. Armed assault is the second most common attack type, with around 40,000 incidents. The regions that make up a large part of the bar are again, Sub-Saharan Africa and South Asia, but also Sub-Saharan Africa. This might be linked to the many active militant groups in these regions.

The least frequently used attack types are hijacking and hostage-taking (barricade incidents). This might be because hijacking and hostage-taking attack types are only popular in specific regions, such as occurring more prominently in North America and Western Europe. These regions may be targeted due to their high-profile visibility and the potential for significant media attention. On the other hand, facility/infrastructure attacks are moderately distributed across multiple regions. This might mean that facility/infrastructure attacks are not region-specific but rather depend on strategic opportunities.

## 
##  Pearson's Chi-squared test
## 
## data:  table(terrorism$success, terrorism$country)
## X-squared = 4579.4, df = 204, p-value < 2.2e-16

We observed the different attack types that occur in different regions. It will be nice to learn if there is a relationship between the different regions and the success of the attack. To test this, we performed a Chi-Square test of independence to test whether the success of an attack is associated with the region where it occurred. The null hypothesis is that there is no association between success and country and the alternative hypothesis is that there is an association between success and region. After running the chi-squared test, we got a p-value of < 2.2e-16. Since the p-value is less than the assumed alpha level of 0.05. This means that we can reject the null hypothesis that there is no association between success and country. We can conclude that success and country are not independent of each other and there is a relationship between them. The heat map below will help us determine how successful each attack type will be based on the region providing more insight into how success rates vary geographically.

The heat map above visualizes the success rate of different attack types across various regions, with the intensity of the color indicating the success rate. Specifically, boxes with a darker blue shade mean higher success rates, while lighter shades indicate lower success rates.

The heat map reveals several important patterns in the success rates of different attack types across regions. For example, unarmed assault in Australasia & Oceania has the lowest success rate, as indicated by the white color. In contrast, hostage-taking (barricade incidents) shows a very high success rate in all regions except North America, where the color is dark purple rather than blue, indicating a slightly lower success rate compared to other regions. Armed assault also shows a high success rate across all regions, as shown by the dark-colored cells. Meanwhile, hijackings in East Asia appear to have the lowest success rate for this attack type, as the cell is noticeably lighter compared to other regions. On the other hand, regions like Sub-Saharan Africa and Central America & Caribbean moderate success rates for most attack types, indicating a less distinct pattern of effectiveness.

Combining the information and analysis for both graphs used in the question, even though bombing/explosion is the most used attack type, it seems to only be modestly effective across different regions, where the color of the boxes is purple which ranges around 0.4 - 0.6 in success. The most used bombing/explosion is in the Middle East & North Africa and South Asia but these regions seem to show modest success (with the color of the two boxes being purple). On the other hand, one of the least frequently used attack types is hostage-taking (barricade incident) ends up being the attack type with the highest success rate. These insights show that counterterrorism groups need to assess not only the frequency of attack types but also their effectiveness when creating strategies to prevent terrorism.

Conclusion

Overall, this report explores various aspects of terrorism patterns using the Global Terrorism Database, focusing on the success rates of attacks, regional casualty trends, and the relationship between attack types against other variables. We focused on four main research questions in our report: what are the key themes and variables to consider when analyzing patterns and trends in this dataset; what are the trends in the number of casualties across different regions over time, and which attack types caused the most damage; how has the percentage of successful terrorist attacks changed for the most active terrorist groups globally, and are certain attack types more likely to succeed in specific regions?

By looking at the wordcloud, we were able to figure out what are the most frequent words like “attack”, “casualties”, and specific terms such as “explosion”, “bomb”, and “fire.” This caused us to focus on variables like attack types, skills (number of casualties for each attack), and their regional impact. We noticed that over time, different regions have different trends in their casualties. For instance, regions like the Middle East & North Africa, Sub-Saharan Africa, and South Asia have had a significant increase in casualties, especially after the 2000s. However, regions like Western Europe and North America experienced stable and low casualties. The attack that has the abnormally highest number of casualties in the Middle East & North Africa is Bombing/explosions. Additionally, the percentage of successful attacks changed by the majority of the most active terrorist groups has a decreasing success rate. This can be because most groups are no longer active. The inactivity can be explained by how the reasons that most groups form are no longer important. Moreover, based on our visualizations and chi-squared test, there is an association between the success and region, and different attack types in different regions have different success rates. For instance, unarmed assault in Australasia & Oceania has the lowest success rate, and hostage-taking (barricade incidents) shows a very high success rate in all regions except North America.

Lastly, some potential future work can be on how geopolitical events influence terrorism patterns like success rate, causalities, etc. This needs future work becuase it requires additional data, such as detailed records of geopolitical events as a variable in our dataset. Even though we did mention some events in our report, these events can not be accurate since we are assuming these are the causes. It would be essential to match terrorism data with their actual events rather than assuming the causation. With our dataset being so big and having so many variables, we can also incorporate advanced machine-learning techniques like classification, random forest, and decision trees into our work. These advanced techniques can help us better capture complex trends that we may have not captured through our current analysis.