Prediction　strategies　based　on　transfer　learning　have　been　proved　to　be　effective　in　solving　Dynamic　Multi-Objective　Optimization　Problems　(DMOPs).　However,　slow　running　speed　impedes　the　development　of　this　method.　To　address　this　issue,　this　paper　proposes　a　transfer　learning　method　based　on　imbalanced　data　classification　and　combines　it　with　a　decomposition-based　multi-objective　optimization　algorithm,　referred　to　as　ICTr-MOEA/D.　This　method　combines　the　prediction　strategies　based　on　transfer　learning　with　knee　points　to　save　computational　resources.　In　order　to　prevent　the　prediction　accuracy　from　being　affected　by　the　insufficient　of　knee　points,　ELM　classifier　selects　more　high-quality　points　from　a　large　number　of　random　solutions.　Moreover,　SMOTE　resolves　the　class　imbalance　problem　during　the　ELM　training　process.　The　simulation　results　demonstrate　that　ICTr-DMOEA　shows　good　competitiveness.　©　2023,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.