Unsupervised Hybrid Algorithm to Detect Anomalies for Predicting Terrorists Attacks

This work presents a hybrid approach for unsupervised algorithms (UHA), in order to extract information and patterns from data concerning terrorist attacks. The reference data are those of the Global Terrorism Database. The work presents an approach based on autoencoders and k-modes type clustering. The results obtained are examined through some metrics presented in the article and it is also considered methodologically how to determine a robust threshold for anomaly detection problems.