Due　to　the　rapid　development　of　the　Internet,　the　data　generated　by　people　in　life　is　growing　at　an　exponential　rate,　and　Short　Message　Service　(SMS)　data　is　one　of　the　social　media　products　of　mobile　phone　users.　The　main　discussion　direction　of　this　article　is　how　to　distinguish　spam　messages　and　obtain　effective　information　from　them,　and　distinguish　the　information　expressed　by　the　messages　themselves.　In　recent　years,　Deep　learning　has　made　great　breakthroughs　in　the　field　of　textual　classification　of　natural　language　processing,　so　this　paper　will　do　more　research　and　breakthrough　on　spam　texts　using　the　deep　learning　method.　This　article　will　introduce　a　Model　(RCM)　combined　with　a　priori　information　extracted　by　Long　Short-Term　Memory　(LSTM)　and　Convolutional　Neural　Network　(CNN).　Firstly,　obtain　the　temporal　and　spatial　information　characteristics　of　the　sentence　using　the　bidirectional　LSTM　as　the　prior　information　of　the　information　itself　before　reading　the　information,　and　then　merge　with　the　feature　information　processed　by　the　CNN　to　improve　the　accuracy　of　the　spam　prediction.　Compared　with　the　previous　model,　there　is　some　innovation　in　getting　the　priori　information　extracted　by　LSTM,　and　it　is　also　improve　the　performance　standard.　©　2018　IEEE.