Application of the KMeans method for clustering the motives of terrorist attacks: a study of data density and distribution

Relevance . Terrorism represents a significant threat to international security, with the diversity and complexity of terrorist attack motives complicating comprehensive analysis. The application of machine learning methods, particularly clustering, allows to identify hidden patterns and trends, providing deeper insights into the underlying causes and preconditions of terrorism. Objective . The study aims to apply the KMeans method to cluster terrorist attack motives, assess the density distribution of data within the identified clusters, and reveal key patterns in the motivational structure of terrorist activities. Methods . The research comprises a dataset of terrorist attack motives extracted from the Global Terrorism Database (GTD), containing descriptions of 209,707 terrorist incidents. The analysis included 22,623 records with clearly defined motives, clustered into 10 groups using the KMeans method. The density and distribution of motives were evaluated using quantitative indicators allowing to assess data concentration within each cluster. Results and analysis. The clustering process allowed to identify 10 clusters of different density and motive structure. The largest cluster (Cluster 0, 12,530 records) is characterized by protest and anti–government motives. Clusters 2, 7, and 9 demonstrated significant data concentration, reflecting distinct motivational trends predominantly related to political and religious extremism. Cluster 7, characterized by sectarian violence motives, exhibited the highest density. Economic motives were significantly underrepresented compared to political and religious ones. Conclusion . The findings confirm the effectiveness of the KMeans method for identifying key terrorist attack motives, despite the complexity and heterogeneity of the initial data. A comprehensive approach to develop preventive measures and counter–terrorism strategies is therefore a collateral.