The　classification　of　municipal　household　waste　is　a　complex,　adaptive　system　containing　multiple　agents,　where　behavior　between　the　agents　is　interactive.　This　study　focuses　on　the　willingness　of　bounded,　rational　individuals　to　sort　waste　within　complex　social　networks,　and　the　evolution　of　group　behavior.　An　individual　cognition-behavior　model　was　established,　based　on　the　theory　of　planned　behavior.　In　this　model,　the　intention　underpinning　individual　waste　classification　is　influenced　by　the　behavioral　attitude　of　the　individual,　perceived　behavioral　control,　subjective　norms,　and　perceived　behavioral　results.　A　multi-agent-based　simulation　within　a　social　network　based　on　the　opinion　propagation　dynamics　and　a　scale-free　network　was　developed　to　simulate　the　behavior　change　of　the　agents.　The　results　suggest　that　under　the　principle　of　selective　connection　and　the　hypothesis　of　network　growth,　the　clustering　trend　of　the　whole　network　towards　several　central　nodes　becomes　increasingly　apparent;　this　means,　the　degree　of　network　node　centralization　becomes　clear.　Lowering　the　threshold　of　social　information　interaction　improved　the　transmission　of　waste　classification　information　and　increased　the　proportion　of　urban　residents　who　carry　out　waste　classification.　Economic　profits　and　spiritual　satisfaction　were　considered　important　driving　factors　that　impacted　decision-making　on　waste　classification,　and　effectively　stimulated　the　willingness　of　urban　residents　to　implement　waste　sorting　behavior.　Notably,　official　organizations　and　institutions,　that　is,　influencers　with　higher　connectivity　were　found　to　have　a　greater　influence　on　the　decisions　made　by　residents.　These　findings　are　useful　as　a　theoretical　reference　to　inform　the　development　of　waste　management　policies.　©　2021