After　＂9.11＂　terrorist　attacks,　more　advanced　information　technologies　have　been　developed　to　counter　terrorism　domain　to　enhance　the　performance　of　early　warning　system.　Machine　learning　based　data　mining　can　be　applied　to　predict　terrorist　event　hidden　in　terrorist　attack　events　and　by　which　the　experts　expect　to　get　a　clear　picture　of　what　the　terrorists　are　thinking　about　in　order　to　step　up　defense　against　these　organized　acts.　This　paper　focuses　on　the　prediction　of　terrorist　event　from　the　Global　Terrorism　Database　(GTD)　with　data　mining　techniques.　Support　Vector　Machine　(SVM),　Naive　Bayes　(NB)　and　Logistic　Regression　(LR)　are　adopted　in　this　paper.　Two　feature　selection　methods　including　Minimal-redundancy　maximal-relevancy　(mRMR)　and　Maximal　relevance　(Max-Relevance)　are　used　to　further　improve　the　classification　accuracy.　Finally,　a　detailed　comparison　of　classification　performance　is　presented,　where　classifier　LR　with　seven　optimal　feature　subset　reaches　a　classification　precision　of　78.41%,　which　validates　the　feasibility　of　applying　machine　learning　to　the　terrorism　field.　We　have　stressed　that　the　classification　methods　can　be　used　to　map　different　inherent　types　in　terrorism　with　both　high　accuracy　and　fast　speed.　Besides,　a　well-chosen　feature　selection　can　lead　to　the　decrement　of　classification　error.