Currently,　the　conventional　deep-ensemble　modeling　approach　faces　several　challenges,　including　intricate　computations,　diminished　accuracy,　and　inadequate　generalization　performance.　In　response,　this　study　introduces　the　Deep　Bayesian　Forest　Fusion　Ensembled　(DBFE)　modeling　method.　Firstly,　a　comprehensive　model　comprising　layered　forests　is　devised,　utilizing　binary　trees　renowned　for　their　strong　interpretability.　Concurrently,　Bayesian　inference　is　employed　as　a　performance　screening　criterion　to　evaluate　the　fitness　of　the　individual-layer　forest　models,　yielding　the　layered　forest　model.　Subsequently,　the　Mahalanobis　distance　between　the　single-layer　predictions　and　the　ground　truth　is　computed　to　ascertain　adherence　to　predefined　disparity　thresholds.　Should　these　thresholds　be　satisfied,　the　reconstituted　dataset　is　forwarded　to　the　subsequent　layer;　conversely,　construction　ceases　and　output　is　generated.　Finally,　leveraging　the　weights　of　the　sub-models,　the　impact　of　each　sub-model＇s　predictions　on　the　ensemble　model　prediction　is　assessed,　thereby　effectuating　the　ensemble　model＇s　prediction.　The　efficacy　of　the　proposed　method　is　corroborated　through　validation　against　benchmark　datasets.　©　2024　IEEE.