The statistical analysis based on GTD terrorist incident record data

Abstract

Terrorism refers to the claims and acts that create social panic, endanger public security, infringe upon persons and property, or threaten state organs and organizations to achieve political and ideological goals through violence, sabotage, intimidation and other means. According to the survey, 92 % of the world's countries have been subjected to terrorist attacks to varying degrees, involving South America, the Middle East and North Africa. Since 1970, there have been about 200,000 terrorist incidents worldwide. Global terrorist attacks have a wide range of targets, involving nearly 20 groups, resulting in up to 280,000 deaths and 360,000 injuries. Based on the GTD database, this paper selects the relevant data from 1998 to 2017, and uses the factor analysis model to establish the evaluation index system of terrorist attack harm from the perspective of terrorist attack harm. At the same time, GIS is used for quantitative analysis of the data, and time series model is used for correlation prediction. The results show that the comprehensive evaluation model of terrorist attacks has objectively and accurately evaluated the terrorist attacks in the past 20 years, among which 9/11 incident has the highest degree of harm, followed by the Salam Embassy bombing. The ranking of the harmfulness of the incident is basically consistent with the reports of mainstream media at home and abroad. From the perspective of the global spatial scope of terrorist attacks, terrorist attacks are shifting from North Africa and the Arabian Peninsula to the eastern front, and the Middle East has continued to gather, and the situation of terrorist attacks in the Middle East is not optimistic. At present and for some time to come, international terrorism will remain at a high level, and terrorist attacks will remain concentrated in the Middle East and the Arabian Peninsula.

1. Introduction

Terrorism refers to the advocacy and behavior of creating social panic, endangering public safety, infringing on personal and property, or threatening state organs and organizations to achieve their political, ideological and other goals through violence, destruction, intimidation and other means. The implementer uses violence or threats of violence against unarmed individuals to achieve their political goals. Since the 1990s, terrorist activities have become increasingly serious, with a series of violent terrorist events represented by the 9/11 attacks not only posing serious threats to the life and property safety of people around the world, but also causing huge economic losses. They have become the primary threat to world stability and regional security.

1.1. Literature review

Current research on terrorism can be broadly divided into two categories: (1) Conduct qualitative or descriptive analysis around the background and consequences of terrorism. (2) Starting from the data of terrorist attacks, using statistical and mathematical methods for research and analysis.

Zhang [1] found through statistical analysis that 92 % of countries worldwide have been subjected to terrorist attacks to varying degrees, including South America, the Middle East and North Africa, Western Europe, South Asia, Central America and the Caribbean, sub Saharan Africa, Southeast Asia, North America, former Soviet countries, Eastern Europe, East Asia, Australasia, Oceania, and Central Asia. At the same time, global terrorist events target a wide range of targets, involving nearly 20 categories. Terrorist attacks kill as many as 280,000 people (6000 per year) and injure as many as 360,000 (8000 per year). Therefore, as an important component of counter-terrorism activities, the threat assessment of terrorist attacks is the basis for effective warning, control, and handling of terrorist attacks [2] GTD (Global Terrorism Database), as the world's most comprehensive open-source terrorism database, collects information on terrorist incidents that have occurred worldwide since 1970, and has now reached around 180,000 cases. The GTD database is planned, maintained, and operated by the University of Maryland with government funding, and is also the START (The National Consortium for the Study of Terrorism and Responses to Terrorism) database [3].

Tang [4] used GTD data to conduct a quantitative analysis of terrorist activities in China, and obtained the evolution patterns and impacts of the surrounding anti-terrorism security risk environment. They believed that the surrounding anti-terrorism security risk environment can be divided into controllable period, active period, relatively stable period, and complex period, and is currently in the most complex stage. Wei [5] used Bayesian network methods to comprehensively evaluate the threats faced by possible terrorist attacks by examining the possible consequences of terrorist attacks, providing decision-making support for counter-terrorism decision-makers to reduce the impact of terrorist attacks. Wang [6] believed that quantitative analysis and research on counter-terrorism issues was an important component of the counter-terrorism research system. Qiao [7] believed that scientifically building anti-terrorism forces and organizing anti-terrorism think tanks in a reasonable manner was an important way to effectively curb and combat terrorist crimes. And considered involving professional research institutions in research, regularly providing anti-terrorism intelligence reports, and dynamically assessing the risks of terrorist activities. Zhu [8] believed through empirical and modeling of current terrorism that its distribution follows a power-law form, and this distribution form was not affected by attack methods and countries, and has strong robustness. Then, empirical statistics were conducted on the time intervals of terrorist incidents, and the results showed that the distribution of time intervals of terrorist incidents also follows a power-law form. Gong [9] established a big data analysis model for comprehensive evaluation of terrorist attack risk prediction, and the results showed that the terrorist attack risk prediction method based on big data analysis has high accuracy and efficiency. Yang [10] proposed the idea of classifying and grading sudden events from a systematic perspective, closely linking the classification and grading of sudden events with the degree of urgency and resource guarantee. In terms of research methods, cluster analysis and discriminant analysis were introduced into the evaluation of classification and grading systems.

In summary, terrorist threat assessment and prediction are current research hotspots and themes. However, there is little literature on establishing an indicator system from the perspective of the harmfulness of terrorist attacks and studying the process of contemporary terrorist attacks from the perspective of spatial evolution. Therefore, based on the GTD database, this paper establishes a comprehensive evaluation index system from the perspective of the harmfulness and influencing factors of terrorist attacks to evaluate the degree of harm caused by terrorist attacks. And use GIS (Geographic Information System) to establish a spatiotemporal evolution feature model, analyze the level distribution, main causes, spread characteristics, spatiotemporal characteristics and other influencing factors of terrorist attacks, and predict the future anti-terrorism situation in key global regions.

1.2. Data source

This paper uses the Global Terrorism Database (GTD) as a raw dataset to obtain detailed information about the attacks. This open source database records more than 200,000 terrorist attacks from 1970 to 2019, with each record containing 135 indicators such as incident information, location, time, weapon information, number of victims, casualties, and property damage. According to the data type, the indicators describing terrorist attacks in GTD are divided into numerical indicators, such as the number of murderers and the total number of deaths; Category-based indicators, such as weapon type, attack type; and text-based indicators, such as weapon details, name of criminal group, etc.

1.3. Research questions and objectives

From a global perspective, almost every day there are terrorist attacks, some of which are extremely bad in nature, and some of which have slightly less impact. If all of them are mixed up in anti-terrorism prevention, I am afraid it will not be effective. Therefore, this paper uses the confirmatory factor analysis method to construct evaluation indicators, score the harmfulness of terrorist attacks, and judge the harmfulness of terrorist attacks according to the score. It makes the judgment of the harm of terrorist attacks more scientific. By using the global spatial autocorrelation index and ArcGIS tool, the spatial evolution characteristics of the harm of terrorist attacks are analyzed. Explore the underlying reasons behind the terrorist attacks.

2. Comprehensive evaluation index system for the harmfulness of terrorist attacks

2.1. Establishment of indicator system

According to the definition of the harm of terrorist attacks, this paper selects indicators from the aspects of social harm, economic harm and personal harm to describe the harm of terrorist attacks [11]. In order to more directly depict the harm of terrorist events, the data quality of each column in the GTD database from 1998 to 2017 was analyzed, and the value integrity, value repeatability and outlier distribution of each column index were discussed. Then the correlation analysis is used to judge the correlation between the two indicators. Finally, the variables involved in confirmatory factor analysis were selected. The final dataset is 114,184 terrorist attacks and 17 million data variables.

Based on the principles of scientific systematicity, goal orientation, and independent comparability [12], combined with the data characteristics and structure in GTD, information integration was carried out on the variables in the data. Finally, the following indicator system was established through classification and summary: the first level indicator is Event Hazard, and the second level indicators are the Event Location, Event Time, Event Nature, Event Attack Ways, and Event Consequences. The third level indicator layer reflects the explanation of the second level indicators. Use this indicator system to specifically express the harmfulness of terrorist attacks.

Based on the existing data information of GTD, the Event Location is divided into two levels: country and city. The countries where terrorist attacks occurred are divided into developed countries and underdeveloped countries according to Gross Domestic Product (GNP) [13], and the attacked cities are divided into capital, important cities, and small cities according to administrative level and city GDP [14]; The Event Time can be divided into occurrence time and whether it is a continuous event (is the duration). Based on the date of terrorist events, the event time variables in the database have been reclassified according to whether it is a holiday, holiday variables have been added, and events can be classified into continuous events and non continuous events according to the time span of terrorist attacks [15]; The Event Nature is characterized by the target of the attack and the scale of the killer; The Event Attack Ways are expressed by the attack way and weapon type; The Event Consequences are represented by five aspects: number of deaths, number of injuries, amount of property damage, number of abductees, and amount of ransom paid; It should be noted that this paper will no longer express the degree of harm, weapon types, and attack targets in the original order in the GTD database. The new order of attack targets is: government, military, government (diplomacy), non-governmental organizations, airport aircraft, food or water supply, transportation (excluding aviation), commerce, educational institutions, religious figures, public utilities, maritime affairs, police, telecommunications, journalists, violent political parties, terrorists, tourists, and private property of citizens themselves; The new order of attack methods is: bombing/explosion, armed attack, assassination, hijacking, kidnapping, facility/infrastructure attack, roadblock incident, unarmed attack; The new order of weapon types is: nuclear weapons, radioactive weapons, light weapons, explosives, chemical weapons, chemical weapons, biological weapons, incendiary weapons, fake weapons, transportation vehicles, and destructive equipment. Through the above indicators, objectively and comprehensively reflect the entire process of terrorist attacks in the entire region from the perspective of harmfulness. However, due to space limitations, only the secondary and tertiary indicators are shown in Table 1.

Table 1.

Hazard index system for terrorist attacks.

Primary indicators	Secondary indicators	Third level indicators
Harmfulness	Event Location	Country City
	Event Time	Occurrence time Is the duration
	Event Nature	The target of attack The scale of killer
	Event Attack Ways	Attack ways Weapon types
	Event Consequences	Number of deaths Number of injuries Number of abductees Amount of property damage Amount of ransom paid

2.2. Establishment of model

In order to make the evaluation of the harmfulness of terrorist attacks objective and truthful, this paper uses confirmatory factor analysis to determine the weights of each indicator. Based on the determined weights, a model of the evaluation index system for the harmfulness of terrorist attacks is formed [[16], [17], [18]]. The purpose of factor analysis is to synthesize multiple variables into independent factors, and the mathematical model is represented by equation (1):

$$R=\left(\begin{array}{c}{X}_1={\alpha }_{11}{\beta }_1+{\alpha }_{12}{\beta }_2+\dots +{\alpha }_{1m}{\beta }_m+\epsilon \\ X_2={\alpha }_{21}{\beta }_1+{\alpha }_{22}{\beta }_2+\dots +{\alpha }_{2m}{\beta }_m+\epsilon \\ ⋮\\ X_p={\alpha }_{p1}{\beta }_1+{\alpha }_{p2}{\beta }_2+\dots +{\alpha }_{pm}{\beta }_m+\epsilon \end{array}\right)$$ (1)

Among them, $X_1,X_2,\cdots ,X_P$ are $p$ original variables, which are standardized variables with a mean of 0 and a variance of 1, ${\alpha }_{pm}$ is the factor load, indicating that $X_1$ is the relative importance on the $\mathrm{m}$-th common factor. The larger the absolute value of ${\alpha }_{pm}$, the stronger the relationship between the common factor ${\beta }_m$ and the original variable $X_p$, where ${\beta }_1,{\beta }_2,\cdots ,{\beta }_m$ represents $\mathrm{m}$ factor variables, $m<p$, $\epsilon$ is a special factor that represents the part of the original variable that cannot be explained by the factor. In this paper, the indicator layer is divided into two layers. Due to the large number of indicators in the second layer, we use the principal component method to synthesize the indicators and form the first level of comprehensive indicators. The model is represented by equation (2):

$$F_j=\frac{{\alpha }_1{F}_1^{\left(j\right)}+{\alpha }_2{F}_2^{\left(j\right)}+\dots +{\alpha }_k{F}_k^{\left(j\right)}}{{\alpha }_1^{\left(j\right)}+{\alpha }_2^{\left(j\right)}+\dots +{\alpha }_k^{\left(j\right)}}$$ (2)

Here, $F_1^{\left(j\right)},F_2^{\left(j\right)},\cdots ,F_k^{\left(j\right)}$ represents the $K$ factor scores of layer $J$; ${\alpha }_1^{\left(j\right)},{\alpha }_2^{\left(j\right)},\cdots ,{\alpha }_{\mathrm{k}}^{\left(j\right)}$ represents the variance contribution of the $K$-th factor in the $J$-th layer; $F_j$ represents the comprehensive score of the $J$-th layer. Using the scores of each factor as variables and the proportion of variance contribution rate of each factor to the total variance contribution rate of the factors as weights, the comprehensive scores of each city are obtained through weighted aggregation.

This paper comprehensively analyzes the different levels of indicators of the harm of terrorist attacks. Use the size of the $F_j$-value to indicate the level of harm caused by terrorist attacks in various regions around the world. The larger the value of $F_j$, the stronger the harm.

The confirmatory factor analysis used in this paper is: the factor analysis is carried out on the third-level indicators, and the weight of each indicator can be obtained from the variance contribution rate. The principal component of the third level index is obtained according to the weight value ranking, and the weight of the second level index is obtained from the variance contribution rate of factor analysis. Finally, the harm value of the final terrorist attack is calculated according to equation (2).

According to the above indicator system, there are five categories of secondary indicators, namely Event Location, Event Time, Event Nature, Event Attack Ways, and Event Consequences. For the convenience of analysis, their three-level indicators are now named: country $x_1$, city $x_2$, occurrence time $y_1$, and is the duration $y_2$; the target of the attack $z_1$, and the scale of the killer $z_2$; attack ways $h_1$ and weapon types $h_2$;number of deaths $m_1$, number of injuries $m_2$, amount of property damage $m_3$, number of abductees $m_4$, amount of ransom paid $m_5$.

The principal components of the Event Location were analyzed based on two factors: country $x_1$ and city $x_2$. The results obtained are shown in Table 2, and the formula is derived from Table 2: $f_1=x_1*1.021$.

Table 2.

Principal component analysis of event location.

Component	Initial Eigenvalue			Extract Sum of Squares Loading
	Total	Var%	Accumulate%	Total	Var%	Accumulate%
1 2	1.021 0.979	51.031 48.969	51.031 100.000	1.021	51.031	51.031

The principal components of the Event Time were analyzed based on the occurrence time $y_1$ and is the duration $y_2$. The results are shown in Table 3, and the time formula can be obtained as $f_2=y_2*1$.

Table 3.

Principal component analysis of event time.

Component	Initial Eigenvalue			Extract Sum of Squares Loading
	Total	Var%	Accumulate%	Total	Var%	Accumulate%
1	1.000	50.015	50.015	1.00	50.015	50.015
2	1.000	49.985	100.000

The principal components of the Event Nature were analyzed based on the target of the attack $z_1$, and the scale of the killer $z_2$. The results are shown in Table 4, and the time formula can be obtained as $f_3=z_3*1.012$.

Table 4.

Principal component analysis of event nature.

Component	Initial Eigenvalue			Extract Sum of Squares Loading
	Total	Var%	Accumulate%	Total	Var%	Accumulate%
1 2	1.012 0.988	0.592 49.408	50.592 100.000	1.012	50.592	50.592

The principal components of the Event Attack Ways were analyzed based on attack ways $h_1$ and weapon types $h_2$. The results are shown in Table 5, and the time formula can be obtained as $f_4=h_1*1.074$.

Table 5.

Principal component analysis of event attack ways.

Component	Initial Eigenvalue			Extract Sum of Squares Loading
	Total	Var%	Accumulate%	Total	Var%	Accumulate%
1 2	1.074 0.926	53.717 46.283	53.717 100.000	1.074	53.717	53.717

The principal components of the Event Consequences were analyzed based on number of deaths $m_1$, number of injuries $m_2$, amount of property damage $m_3$, number of abductees $m_4$, amount of ransom paid $m_5$. The results are shown in Table 6, and the time formula can be obtained as $f_5=m_1*1.623$.

Table 6.

Principal component analysis of event consequences.

Component	Initial Eigenvalue			Extract Sum of Squares Loading
	Total	Var%	Accumulate%	Total	Var%	Accumulate%
1 2 3	1.623 1.000 0.377	54.093 33.333 12.573	54.093 87.427 100.000	1.623	54.093	54.093

Finally, the above principal components $f_1$ to $f_5$ were subjected to factor analysis, and three common factors were identified. The comprehensive ranking was analyzed by weighting their variance contribution rates (as shown in Table 7), and we obtained equation (3):

$$f=\frac{f_1*1.249+f_2*1.062+f_3*0.9981}{1.249+1.062+0.998}$$ (3)

In this paper, factor analysis is carried out on the third level index, and the weight of each index can be obtained from the variance contribution rate. The principal component of the third level index is obtained according to the weight value ranking, and the weight of the second level index is obtained from the variance contribution rate of factor analysis. Finally, the harm value of the final terrorist attack is calculated according to formula (2).

Table 7.

Principal component analysis of hazard factors.

Component	Initial Eigenvalue			Extract Sum of Squares Loading			Rotational Sum of Squares Loading
	Total	Var%	Accumu-late%	Total	Var%	Accumu-late%	Total	Var%	Accumu-late%
1 2 3 4 5	1.249 1.062 0.998 0.937 0.753	24.989 21.247 19.960 18.738 15.067	24.989 46.235 66.195 84.933 100.000	1.249 1.062 0.998	24.989 21.247 19.960	24.989 46.235 66.195	1.245 1.065 1.000	24.892 21.301 20.003	24.892 46.192 66.195

According to the obtained danger coefficient score, the severity of the harm of terrorist attacks was graded, and the levels were sequentially reduced. The results are shown in Table 8.

Table 8.

Top 10 most hazardous events.

Event Number	Comprehensive score	Notes
200109110005 200109110004	52.4	9/11 attacks
199808070002	52.39	The explosion at the Salaam embassy
201603080001	18.72	Baghdad Suicide Terrorist Attack
201406150063	12.58	The Iraq Massacre
200409010002	9.79	The Beslan hostage incident
200802010006	9.42
200607120001	7.26	Mumbai bombings in India
201710010018	6.83	Las Vegas terrorist attacks
201408090071	6.83
200708150005	6.48

Therefore, by establishing a comprehensive indicator evaluation model that reflects the degree of harm, it can be concluded that this model can objectively and accurately evaluate the scores of terrorist attacks in the GTD database, thereby helping to define the degree of harm of terrorist attacks from subjective judgment to objective evaluation scores.

3. The spatial evolution characteristics of the harmfulness of terrorist attacks

3.1. Spatial distribution model

Using ArcGIS software tools, terrorist attacks that occurred globally in 2015, 2016, and 2017 were classified into 6 levels based on natural breakpoints, ranging from low to high, and classified by color depth. They are divided into six levels (0 times), five levels (0–21 times), four levels (21–83 times), three levels (83–278 times), two levels (278–1244 times), and one level (1244–2750 times), as shown in Fig. 1.

Fig. 1.

Distribution differences of Global

Terrorist Attacks in 2015.

From Fig. 1, Fig. 2, Fig. 3, it can be seen that: (1) In the past three years, there have been 18, 11, and 9 regions that have achieved level 2 or above terrorist attacks, indicating that terrorist attacks have been decreasing year by year on a global scale. (2) The areas where terrorist attacks occurred over the past three years are still concentrated in the Middle East and the Arabian Peninsula. (3) From the perspective of time and regional changes, terrorist attacks are shifting from North Africa and the Arabian Peninsula to the east, and the Middle East has been continuously gathering, indicating that the situation of terrorist attacks in the Middle East is not optimistic [19].

Fig. 2.

Distribution differences of Global

Terrorist Attacks in 2016.

Fig. 3.

Distribution differences of Global Terrorist Attacks in 2017.

3.2. Global spatial relationship model

Measured by the global spatial autocorrelation index ($Moran’sI$), which reflects the degree of spatial correlation and clustering of variables. equation (4) is as follow:

$$I=\frac{n\sum _i\sum _j{w}_{ij}\left(x_i-\bar{x}\right)\left(x_j-\bar{x}\right)}{\left(\sum _i\sum _j{w}_{ij}\right)\sum _i{\left(x_i-\bar{x}\right)}^2}$$ (4)

Among them, $I$ is the global $Moran’sI$ index; $n$ is the total number of countries; $X_i$, $X_j$ represents the number of terrorist attacks suffered by country $i$ and country $j$, $\bar{x}$ represents the average number of terrorist attacks suffered by a country in a certain region in a certain year. $W_{ij}$ is the spatial weight matrix. When regions $i$ and $j$ are adjacent, the $W_{ij}$ value is 1, otherwise $W_{ij}=0$, and the range of $I$ values is [−1,1]. When $I>0$, it indicates a spatial positive correlation between a country's terrorist attacks. Conversely, if $I<0$, it indicates a spatial negative correlation between a country's terrorist attacks. The occurrence of terrorist attacks tends to be spatially discrete. If $I=0$, it indicates that terrorist attacks occurring in that region are randomly distributed [20]. The significance level of spatial data can be determined based on its distribution, and its significance formula is represented by equation (5):

$$Z\left(I\right)=\frac{Moran\prime sI-E\left(Moran\prime sI\right)}{\sqrt{VAR\left(Mora{n}^{\prime }sI\right)}}$$ (5)

Table 9 shows the indices and significance test values for measuring the spatial autocorrelation and clustering $Moran’sI$ of countries affected by terrorist attacks worldwide, calculated using (4), (5). Among them, the normal statistical value Z of the $Moran’sI$ index in the table is greater than the critical value of the normal distribution function at the 0.01 level (1.96), indicating that the spatial distribution of global terrorist attacks has a clear spatial dependence, and also indicating that the number of global terrorist attacks does not show a completely random state in spatial distribution, but rather shows spatial clustering between similar values, Positive spatial correlation represents similar characteristics of neighboring countries and regions, that is, countries and regions with more attacks tend to be closer to those with more attacks, or countries and regions with lower attacks tend to be closer to those with lower attacks. Therefore, overall, there is spatial correlation and significant spatial clustering among countries and regions where terrorist attacks occur globally [21].

Table 9.

Moran's I index and its Z values for global terrorist attacks from 2015 to 2017.

Year	Moran's I	Z Value	P Value
2015	0.148057	2.5305	0.0000
2016	0.110263	2.5211	0.0000
2017	0.111394	2.5279	0.0000

From Fig. 4, Fig. 5, Fig. 6, it can be seen that: (1) the number of HH regions increased from 7 in 2015 to 8 in 2016, and then decreased to 7 in 2017, indicating that the spatial aggregation ability of terrorist attacks first increased and then weakened, demonstrating the persistence of terrorist attack areas. (2) The number of terrorist attacks in the LH region has been decreasing year by year, from 12 in 2015 to 10 in 2016, and then to 8 in 2017.

Fig. 4.

LISA map of global terrorist attacks in 2015.

Fig. 5.

LISA map of global terrorist attacks in 2016.

Fig. 6.

LISA map of global terrorist attacks in 2017.

In summary, new ‘Arcs of Turbulence’ [22] have emerged in the vast regions of Central Asia, northern South Asia, West Asia and North Africa, as well as the Sahel region. On this ‘Arcs of Turbulence’, international terrorism is both independent and interconnected [23]. In the “Arc of Turbulence” region, international terrorism carries out terrorist activities along geopolitical fragmentation zones, geopolitical economic competition zones, and geopolitical Islamic distribution zones [24]. The coexistence, interdependence, and collusion between international terrorism and domestic terrorism are the most prominent characteristics of terrorism in the world today, and they are also an important reason for the rampant spread of international terrorism.

Central Asia, northern South Asia, West Asia, North Africa and the Sahel region have developed into a multi-ethnic, multi-religious and multi-cultural complex due to their geographical characteristics and historical origins. In the process of social development, there has been a wide gap between the rich and the poor and serious ethnic antagonism, which has become one of the incentives for terrorist attacks in these regions. Most of the terrorist groups in these regions have religious backgrounds and advocate a return to Islamic traditions, such as Indonesia's Jemaah Islamiyah (JI) and militia Jihad (LJ). The opposition of religious belief caused the separation of ethnic relations and intensified the formation of conflicts. Religious extremism and secession are not two separate paths, and the combination of the two leads to more serious terrorist attacks.

Currently and in the foreseeable future, international terrorism will continue to maintain a high incidence due to the uncertain and turbulent international and regional situation. Although the extremist organization ISIL has been heavily damaged, it is difficult to completely eradicate it in the short term. After losing power, the Islamic State will transform into a normalized terrorist organization. For a considerable period of time, the development of global terrorism will exhibit characteristics such as globalization, homogenization, socialization, and networking [25].

The current international terrorist activities are constantly showing a trend of diversification, and the path of terrorism is a political process, with political confrontation as its internal factor [26]. With the evolution and variation of terrorist attacks, especially the emergence of new organizational forms such as “lone wolves”, local terrorists, and ISIS, it is necessary to go beyond conceptual models such as “terrorist organizations”, “non-traditional security”, and “asymmetric confrontation” to re-examine the threat of contemporary terrorism.

4. Recommendations

According to the characteristics and trend of terrorist attacks, all countries in the world should strengthen international counter-terrorism cooperation and internal counter-terrorism struggle to stabilize the security situation in various regions.

• (1)Strengthen political mutual trust and reach consensus on counter-terrorism under the guidance of a community of common destiny and the common interests of all countries;
• (2)Establish a larger counter-terrorism cooperation organization with a wider scope of cooperation, establish permanent offices, and build a regular counter-terrorism cooperation mechanism;
• (3)Establish a database of counter-terrorism intelligence information around counter-terrorism cooperation organizations, increase investment in the collection, analysis and mining of counter-terrorism intelligence information, and give priority to intelligence information sharing;
• (4)Coordinate and provide support for counter-terrorism capacity building, joint counter-terrorism exercises and training of counter-terrorism personnel;
• (5)Carry out in-depth cooperation on the identification, investigation, evidence collection, extradition and repatriation of transnational terrorists, and remove international obstacles.
• (6)Improve the construction of border isolation facilities, regularly carry out special border prevention and control operations, crack down on all kinds of illegal and criminal activities, and maintain border security at all times.

5. Conclusion

Based on GTD database, this paper establishes a comprehensive evaluation system of terrorist attacks by using confirmatory factor analysis. Using GIS combined with statistical analysis method, quantitative analysis of data and related prediction of future terrorist attacks were carried out. The results are as follows:

The comprehensive evaluation model of terrorist attacks has objectively and accurately evaluated the terrorist attacks in the past 20 years, among which 9/11 is the most harmful. It was followed by the embassy bombing in Salam, the suicide terrorist attack in Baghdad, the massacre in Iraq, the hostage incident in Beslan, the serial bombings in Mumbai, India, and the terrorist attack in Las Vegas, and the ranking of the harm of the events was basically consistent with the reports of mainstream media at home and abroad. From the perspective of the global spatial scope of terrorist attacks, terrorist attacks are shifting from North Africa and the Arabian Peninsula to the eastern front, and the Middle East has continued to gather, and the situation of terrorist attacks in the Middle East is not optimistic.

The symbiotic relationship between international terrorism and domestic terrorism is the most prominent feature of today's terrorism, and it is also an important reason why international terrorism cannot survive a fight and is rampant. At present and for some time to come, international terrorism will continue to maintain a high incidence and diversified situation. At the same time, from the perspective of the lack of spatial spread, the development trend of international terrorism has gradually increased the pressure on China to fight terrorism both inside and outside. In particular, the anti-terrorism situation in the surrounding areas has become increasingly severe, and the threat to overseas security interests has risen sharply. It is crucial to expand international counter-terrorism cooperation, especially with major countries, neighboring countries and key regional countries, which will also help ease the pressure on all countries to fight terrorism. This conclusion accords with the actual situation of the place where the terrorist attack occurred.

6. Discussion

This paper makes use of GTD database to explore and analyze the data of the terrorist attacks that have occurred, but the GTD database is not fully utilized. The following questions are worth thinking and studying.

• (1)Confirmatory factor analysis has been widely used in the analysis of terrorist attacks, but how to combine with machine learning models to predict future trends still needs further discussion;
• (2)The analysis of terrorist attacks in this paper ignores the impact of the external environment on terrorist organizations, and only starts from the data itself. However, in reality, terrorist attacks and political environment and other unquantifiable factors are closely related, which will also lead to certain errors between the analysis and prediction results of this paper and reality.

Availability of data and material

The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.

CRediT authorship contribution statement

Yunhuan Qu: Writing – original draft, Investigation. Yatian Chen: Writing – original draft, Resources, Project administration. Zhaowen Tan: Writing – review & editing, Data curation. Bei Han: Writing – review & editing, Software.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.