Complex　event　detection　technology　aims　at　extracting　valuable　information　from　continuously　incoming　massive　stream　data　quickly　and　accurately,　which　is　one　of　the　key　components　of　complex　event　processing　platform.　XML　is　usually　regarded　as　one　of　the　data　models　of　complex　event　processing　platform.　Although　there　are　many　efficient　filtering　algorithms　for　XML　stream　data,　they　are　generally　unable　to　support　complex　event　queries　that　contain　timing　relationships.　So　we　used　the　regular　tree　pattern　to　describe　the　complex　event　pattern,　generated　the　macro　forest　transducers　and　the　automatic　for　regular　part　matching　based　on　the　regular　tree　pattern　grammar　and　proposed　an　efficient　complex　event　detection　method.　The　experimental　results　prove　that　the　method　have　strong　capability　on　complex　event　detection.