머신러닝을 활용한 테러리즘 빅데이터 분석

Abstract

Listen

Translate article

The purpose of this study is to analyze terrorism-related structured and unstructured big data using machine learning techniques for big data analysis to confirm the predictability of terrorism and derive implications. To this end, big data on terrorism was obtained from research institutes specializing in terrorism, articles and documents posted on various portals and SNS were collected, put into machine learning, and analyzed and learned, and the results were confirmed. The analysis results are as follows. First, as a result of conducting K-Means Clustering, a representative unsupervised learning among machine learning techniques, eight regional clusters of terrorism occurred around the world in 2021 (near Afghanistan, Democratic Republic of the Congo, India, Iraq, Myanmar, Nigeria, Pakistan, and Somalia). Next, as a result of clustering by type of terrorism, assassination using weapons excluding explosives, kidnapping or kidnapping targeting key facilities were identified. In addition, as a result of clustering around the means used in terrorism, it was confirmed that there were many explosives ahead, but there were differences according to the cluster. Finally, clusters of military, police, government personnel, civilians, and private property were identified as a result of clustering centered on targets of terrorism. Second, as a result of sentiment analysis by setting unstructured texts on terrorism events as research subjects, it was confirmed that there were differences in views on terrorism between terrorism researchers and the general public. Based on these research results, policy suggestions were made on ways to prevent terrorism.