Machine Learning Model for Predicting Terrorist Attacks in Sub-Saharan Africa

Abstract

Security of lives and properties is an essential aspect of governance which any serious government of a nation must not toy with. The assurance of safety of the populace however is not to be left in the hands of the government alone. If adequate security is to be achieved, there must be participation from corporate bodies, non-governmental organisation, communities, and even private setup. The Sub-Saharan Africa is probably becoming a den of terrorists. The operations and activities of the terrorists in the Sub-Saharan are so volatile that their next possible move are difficult to predict. To this end, this study is poised to establish machine learning models capable of predicting the terrorists attack in Sub-Saharan African by relying on the past reservoir of attacks data on the continent. The study which considers the peculiarity of attacks in the Sub-Saharan region obtained past data from the Global Terrorism Database, spanning from 1970 to 2017 with data on Sub-Saharan African majorly utilized. The approach adopted sees the study deploying Random Forest Method and the Long-Short Term Memory (LSTM) Technique. The results show that the two methods deployed are capable of predicting terrorist attacks in Sub-Saharan African with prediction accuracy of 85.1% and 80.1% respectively. In the nearest future, it is hoped that this study will be further researched in order to evolve into a data-driven system that predicts with actual data and information, the next probable attacks of the terrorists in terms of time and place.