Deep analysis of the data related to terrorist attacks will help people to have deeper understanding of terrorism and provide valuable information supporting for counter-terrorism and anti-terrorism. Based on the global terrorism database, this paper introduces artificial intelligence theory, such as machine learning to conduct data mining and quantitative analysis of terrorist attack records. A quantitative classification model based on D-S evidence theory and adaptive Gauss Cloud Transform (AGCT) algorithm is constructed. Firstly, attribute dimensionality is reduced with expert knowledge and correlation analysis, and significant hazard assessment indicators are selected. Finally, based on the AGCT algorithm, the hazard index is classified to determine the hazard level of the event.