An ensemble machine learning model for the prediction of danger zones: Towards a global counter-terrorism

Abstract

Terrorism can be described as the use of violence against persons or properties to intimidate or coerce a government or its citizens to some certain political or social objectives. It is a global problem which has led to loss of lives and properties and known to have negative impacts on tourism and global economy. Terrorism has also been associated with high level of insecurity and most nations of the world are interested in any research efforts that can reduce its menace. Most of the research efforts on terrorism have focused on measures to fight terrorism or how to reduce the activities of terrorists but there are limited efforts on terrorism prediction. The aim of this work is to develop an ensemble machine learning model which combines Support Vector Machine and K-Nearest Neighbor for prediction of continents susceptible to terrorism. Data was obtained from Global Terrorism Database and data preprocessing included data cleaning and dimensionality reduction. Two feature selection techniques, Chi-squared, Information Gain and a hybrid of both were applied to the dataset before modeling. Ensemble machine learning models were then constructed and applied on the selected features. Chi-squared, Information Gain and the hybrid-based features produced an accuracy of 94.17%, 97.34% and 97.81% respectively at predicting danger zones with respective sensitivity scores of 82.3%, 88.7% and 92.2% and specificity scores of 98%, 90.5% and 99.67% respectively. These imply that the hybrid-based selected features produced the best results among the feature selection techniques at predicting terrorism locations. Our results show that ensemble machine learning model can accurately predict terrorism locations.