With　the　development　of　information　technology　and　communication　industry,　there　is　a　phenomenon　of　malicious　arrears　of　some　telecom　users,　which　brings　economic　losses　to　telecom　operators.　To　solve　the　above　problems,　this　paper　proposes　a　credit　risk　prediction　method　for　telecom　users　based　on　model　fusion.　We　utilized　three　popular　and　diverse　classifiers　as　the　individual　learner,　including　k-nearest　neighbor　(KNN),　random　forest　and　extreme　gradient　boosting　(XGBoost),　the　results　of　the　three　classifiers　were　then　fitted　using　logistic　regression.　To　verify　this　model,　real　data　from　Chinese　telecom　users　were　used　in　a　test.　According　to　the　experimental　results,　the　proposed　fusion　model　significantly　outperforms　the　compared　methods　in　terms　of　recall　rate,　F1　value　and　AUC　value,　which　can　effectively　predict　the　credit　risk　of　telecom　users.　©　2021　IEEE.