Tricking　account　overdraft　fee　refers　to　the　behavior　of　the　users　who　utilize　the　overdraft　limit　to　purchase　goods　but　no　longer　recharge　the　account.　In　order　to　solve　the　problem　that　some　mobile　communication　users　have　a　tricking　account　overdraft　fee　behavior　which　lead　to　bad　debts　of　telecom　operators,　a　risk　prediction　model　based　on　the　combination　of　Logistic　and　GBDT　is　proposed.　The　fusion　model　uses　Logistic　Regression　transformation　to　map　the　function　value　to　the　interval　from　0　to　1,　and　the　mapped　value　was　used　as　the　risk　probability　value　for　prediction;　GBDT　is　used　to　discover　features　with　multiple　degrees　of　differentiation　and　combine　effective　features　to　enhance　the　feature　dimension.　The　combination　features　were　extracted　from　the　original　data,　which　made　up　for　Logistic　Regression＇s　poor　capture　of　feature　combinations　and　improved　the　predictive　ability　of　this　model.　The　experiment　based　on　real　mobile　communication　user　data　of　an　operator　shows　that　the　proposed　fusion　model　has　a　good　prediction　effect.