The　results　of　the　existing　research　on　ensemble　methods　based　on　resampling　are　the　best　for　the　imbalance　classification　task,　whereas　the　results　of　independent　use　of　resampling　or　ensemble　learning　are　relatively　mediocre.　Compared　with　the　other　two　ensemble　strategies　such　as　boosting　and　bagging,　stacking　is　often　better　than　iterative　boosting　in　training　speed　because　it　is　an　ensemble　strategy　that　parallels　base　classifiers.　Bagging　uses　a　weighted　average　at　the　decision　level,　whereas　stacking　uses　a　machine　learning　model　to　make　decisions,　which　achieves　higher　accuracy　than　bagging　using　a　weighted　average　decision.　The　effect　of　stacking　relies　heavily　on　the　variance　of　the　base　classifiers,　and　it　is　often　better　to　ensure　that　the　base　classifiers　have　good　learning　performance　while　making　the　variance　among　base　classifiers　as　large　as　possible.　The　key　to　imbalanced　classification　is　to　deal　with　the　class　imbalance　and　overlapping　in　training　sets.　In　this　study,　we　will　use　a　hierarchical　approach　to　solve　these　two　problems.　Thus,　we　propose　a　resampling-based　stacking　ensemble　method　combined　with　clustering.　The　proposed　method　uses　a　support　vector　machine　as　the　base　classifier　to　solve　the　problem　of　losing　important　information　and　easily　introducing　noise　in　the　resampling　process　through　clustering.　It　also　increases　the　differences　among　base　classifiers.　We　use　the　proposed　stacking　ensemble　method　to　hierarchically　solve　the　class　overlapping　and　then　use　the　cost-sensitive　algorithm　to　solve　the　class　imbalance.　The　statistical　significance　analysis　of　the　experimental　results　shows　that　the　proposed　method　performs　better　than　existing　imbalanced　ensemble　methods.　©　2022　IEEE.