Predicting the Success of Global Terrorist Attacks Based on Machine Learning

Terrorist attacks can be analysed and predicted with detailed historical data for better prevention and early warning. In this paper, we use the Global Terrorism Database (GTD) to quantitatively study the characteristics of terrorist attacks from 1970 to 2019, and construct a framework based on four machine learning classification algorithms, namely, decision trees, random forests, logistic regression, and XGBoost, to predict whether a terrorist attack can be successful. A skillful method is used to deal with the unbalanced data. Each classifier undergoes optimizing, training, and parameter search to attain optimality. The proposed framework is then evaluated to assess its performance and stability. The experimental results show that the four algorithms achieved good performance. Random Forest and XGBoost achieve the best prediction performance of high accuracy (99%) in predicting the success of terrorist attacks.