Accurate　traffic　flow　prediction　is　of　paramount　importance.　Unlike　predictions　centred　on　individual　intersections,　the　complexity　and　interconnectedness　of　traffic　flows　within　a　road　network　pose　unique　challenges　to　traffic　flow　prediction.　Furthermore,　the　traditional　focus　on　steady-state　factors　of　traffic　flow　alone　is　insufficient,　given　the　significant　impact　of　non-stationary　factors　on　traffic　dynamics.　To　address　these　intricacies,　this　study　introduces　a　road　network　traffic　flow　prediction　model,　BF-SAGE-GRU,　which　integrates　the　Butterworth　filter,　GraphSAGE　and　GRU.　The　model　transform　the　traditional　single-intersection　prediction　into　a　road　network　prediction.　Subsequently,　the　accuracy　of　BF-SAGE-GRU　is　compared　with　that　of　mature　models　commonly　used　in　traffic　flow　prediction.　The　results　demonstrate　that　BF-SAGE-GRU　has　superior　performance,　thereby　verifying　its　effectiveness　in　the　field　of　traffic　flow　prediction　in　road　networks.　©　2024　Copyright　held　by　the　owner/author(s).