Terrorist attacks In Africa: A Data-Mining Approach
Terrorist attacks in Africa has been an issue of concern for several decades. The intelligence and security agencies across the globe allocate adequate resources to fight the terrorist attacks. The attacks by terrorists have adverse effects such as infringing sovereignty, national demoralization, government instability and strategic damage. Currently, data mining techniques and methods are applied in the association of criminal data and identification of patterns that were initially not put in use. This paper therefore aims to propose how associations in criminal data and identification of patterns can be used to prevent terrorist attacks in Africa. The approach can offer outcomes that are beyond the macro-level studies that utilize regression models and hypothesis testing. From a practical perspective, the methodology can also be used to offer preventive strategies in the prevention of terrorist attacks in Africa through changing of the systematic and physical environment in the African continent. This works in relation to the Situational Crime Prevention theory. Global Terrorism Database will be used in the application of the methodology. This can be done by presenting and analyzing terror attacks with their patterns with their impact on the casualties. This paper proposes differentiating characteristic exploration as way to minimize the terror attack cases in Africa.
With a focus on understanding the logic behind the actions of the terrorists and to come up with more effective strategies for the terror attacks in Africa. There are various terror attacks typologies as suggested by the criminologists. The number of victims being the highest variable explored. There are terror attack components such as suicide attack, explosions, rush hour attacks and coordinated bombings which are regarded as “new” that aims to increase the number of fatalities and injuries (Hoffman, 2016). There are very many devastating consequences caused by large attacks form terrorists. Terrorist attacks in Somalia and Kenya caused by Al-Shabaab and Al-Qaeda in Western countries has various negative impact on the economy of Somalia and loss of life in Kenya. Increased terror attacks in a country scares the potential investors and reduces the Foreign Direct Investment (FDI) (Shahbaz, 2013). Terror attacks also increases insecurity cases which are considered to affect the tourism sector. Other effects of terror attacks include; stress and public fear. Furthermore, terror attack results in political intolerance, military confrontations and sovereignty infringement. The devastating effect of terror attack enhances the government to come up with tools to prevent and assess terror attacks.
As a way of understanding the causality actions of the terrorists, unique characteristics are identified by the scholars. Actions guided by religion are perceived to cause more victims. An average of 38% of terror causalities emanate from religious acts as opposed to about 9% caused by the nationalists or leftists acts according to Terrorism Knowledge database. Ideologies factors are the promoters of terror attacks as suggested by Mierau (2015). The combination of religious and ethno-nationalist ideologies form dangerous terror groups and then followed by the religious groups as proposed by Asal and Rethemeyer (2008). There are several reasons for the engagement of the religious groups in terror attacks and this involves dehumanizing of the terror attack victims in presence of the religious terrorists (Berman, 2016), large scope of the target, spiritual other than practical and perceiving violence as the required goal other than being uses a tool to attain an important thing.
The proposed study will apply the interpretable classification models in the identification of terror attack patterns with regard to the well-known historical data. In this case, a pattern is a set of values representing certain database features. The Global Terrorism Database (GTD) will be used in this proposed study. The GTB is an open source database containing the terroristic attacks across the globe form the 1970 to 2021. University of Maryland’s National Consortium for the Study of Terrorism and Response to Terrorism (START) is the provider of GTD. The database comprises of more than 171,000international and domestic terror attacks which comprise the perpetrators of the attacks, targets, location, event outcomes and perpetrators.
The analysis will be limited on the most recent terror attacks in Africa. There were 10000 cases between the year 2015 and 2021. The proposed study will make an assumption that there have been a change in the occurrence of the terror attacks since 2014 and hence any data before 2014 is regarded as less relevant to the analysis. The aim of the proposed study will be to determine the variations between mass and low casualty attacks and later applies the unlabeled and logic data. Creation of a classifier field helped in identifying the mass-casualty event that will be based on the confirmed number of fatalities, non-fatal injuries, the injured perpetrators and the perpetrator fatalities. For simplification of the analysis, the class filed will be discretized into categories each being represented by a range of values. The aim of the discretization method will be to differentiate the terror cases with no casualties with some other events and to establish the prevalence histogram of various classes in a way that the terror attacks in every category apart from the “none” category was same.
There are recent studies that have utilized the supervised machine learning algorithm in the detection of terror attacks and crimes. For instance, the posts are classified according to their intention in the online hacking platforms, topic modelling is used in the investigation of the association in the full-array formal crime type that is used by the detectives and the texts that are related to crimes. The random forest algorithm can also be used in the detecting of the cyberspace messages in metadata. The aim of this paper is to propose how the data mining algorithms can be used in the identification of the terror incidents which have a common characteristics and which causes major fatalities. The aim of this study is to demonstrate the utility of applying data-mining algorithms to identify subgroups of terrorist incidents that share common characteristics and that result in mass fatalities. The approach can offer outcomes that are beyond the macro-level studies that utilize regression models and hypothesis testing since this approaches identify features which are related to affected people in the entire sample. The study unlike other previous studies will not entirely assume space and time which comprises of multi-causality and low attacks. The patterns are very crucial in the prevention of terror attacks through extraction of actions and rules that allows security personnel and agencies to make a follow-up. This will work in relation to the Situational Crime Prevention theory (SCT). Global Terrorism Database will be used in the application of the methodology. The Situational Crime Prevention theory (SCT) suggests reduction of the crime opportunities other than trying to change or understand the motives and disposition of the terrorists (Clarke, 2010).
The proposed study will evaluate the classifiers performance starting with the “high” category that is regarded as a “positive” event. The category is very crucial since the study will aim at identifying patterns that encompass the mass-casualty terror attacks. AUC was the overall performance measure for this “high” category which is the area under the area below Receiver Operating Characteristics (ROC). The curve will be established through plotting of the True Positive Rate (TPR). The points on the ROC shows the model performance of the classification algorithm (Green, 1966). Furthermore, the Cohen’s Kappa Statistic (KE) and overall AUC will be computed for the five categories. KE refers to a metric used to compare the expected accuracy with the observed accuracy.
This is the last phase of data-mining process. This stage comprises of the organization of the gained knowledge and presenting it in a form that can be utilized by others. After selecting the best classifier, the proposed study will apply the model in the derivation of the practical insights to assist the criminologists to make an exploration of strategies that prevents multi-causality terror attacks.
In Conclusion, the project will be carried out as per the Cross-Industry Standard Process for Data Mining (CRISP-DM) which was suggested by Chapman et al. (2000). CRISP-DM has six stages; Business understanding is first stage which involves understanding the requirements and objectives of the project from a business perspective. Data understanding being the second stage involves learning the features in the database and being able to identify problems that are related to the quality of data. Data preparation stage will involve the preprocessing options like data cleaning, data transformation and reduction. Data cleaning on the other hand will deal with the exceptional and irrelevant data. In the data reduction step, in cases where two features were strongly inter-correlated. In data reduction step, two features that were related led to one feature being removed. The modelling stage will be used in the classification of the multi-casualty terror attacks in Africa and identifying the significant patterns that cause multi-casualty attacks. The Interpretable models like the Bayesian Models and decision trees can be applied in finding the influential parameters to differentiate low and mass casualty terror attacks (Kotsiantis, 2007).
Berman, E., & Laitin, D. (2006). Rational martyrs versus hard targets: Evidence on the tactical use of suicide attacks. In E. Meyerson (Ed.), Suicide bombing from an interdisciplinary perspective. Princeton, NJ: Princeton University Press
Clarke, R. V. (2010). Situational crime prevention: Theoretical background and current practice. In M. D. Krohn, A. J. Lizotte, & G. P. Hall (Eds.), Handbook on crime and deviance. New York: Springer
Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. New York: Wiley
Hoffman, B. (2016). Inside terrorism. New York: Columbia University Press.
Kotsiantis, S. (2007). Supervised learning: A review of classification techniques. Informatica, 31, 249–268
Shahbaz, M., Shabbir, M. S., Malik, M. N., & Wolters, M. E. (2013). An analysis of a causal relationship between economic growth and terrorism in Pakistan. Economic Modeling, 35, 21–29