Inter-organizational　cooperation　in　innovation　has　evolved　into　a　complex　form　featuring　multilayer　network　interactions.　By　collecting　diverse　and　heterogeneous　data　on　cooperative　innovation,　the　paper　constructs　a　complex　multilayer　network　integrating　inter-subject　knowledge-technology　multiplex　cooperation　network　and　subject-knowledge　interdependent　network,　establishes　a　multi-dimensional　indicator　system　used　for　link　prediction,　then　conducts　link　prediction　based　on　machine　learning　algorithm　and　carries　out　empirical　research.　Research　finds　that:　(1)　Random　Forest　has　the　best　prediction　performance,　and　inter-layer　interaction　information　can　improve　the　prediction　accuracy.　Compared　to　behavior　interaction,　the　inter-layer　knowledge　interaction　plays　an　obvious　role　in　prediction,　which　indicates　that　the　linkage　effect　between　knowledge　cooperation　and　technical　cooperation　network　is　not　obvious　yet.　(2)　In　the　prediction　results,　the　regional　changes　of　cooperation　have　shown　a　trend　of　extending　from　intra-provincial　to　inter-regional　cooperation,　from　coastal　to　western　inland.　Besides,　the　strong　association　of　industry-university,　industry-research,　leading　enterprises　and　kinship　cooperation　are　important　cooperation　modes　in　the　future.　©　2023　ACM.