This　study　suggests　a　hybrid　machine　learning　(ML)　and　multi-objective　optimization　(MOO)　methodology　for　improving　decision-making　in　Australian　gold　mining　operations.　Traditional　heuristics　are　poorly　performing　due　to　process　variability,　and　therefore　ML　models　namely　the　CatBoost　regressor—were　employed　to　predict　with　high　accuracy　the　key　performance　indicators　(KPIs)　of　ore　processed,　energy　consumed,　cost,　and　greenhouse　gas　(GHG)　emissions.　CatBoost,　optimized　by　the　Grey　Wolf　Optimizer　(GWO),　yielded　satisfactory　predictive　performance　with　an　R2　of　0.978,　MAE　of　3.361,　and　MAPE　of　0.0745　in　forecasting　GHG　emissions　intensity.　The　hyperparameter-tuned　CatBoost　model　was　utilized　as　the　objective　function　in　a　multi-objective　optimization　framework　using　the　Constrained　Two-Archive　Evolutionary　Algorithm　(C-TAEA).　Six　bi-objective　scenarios　were　examined　on　the　basis　of　production-cost,　production-energy,　production-emissions,　cost-energy,　cost-emissions,　and　energy-emissions　trade-offs.　Among　them,　the　energy　consumption　and　emissions　minimization　scenario　yielded　the　best　hypervolume　outcomes,　representing　Pareto-optimal　solutions　with　high　quality.　Visualization　techniques　like　parallel　coordinate　plots　and　hypervolume　indicators　were　used　to　assess　the　trade-offs,　and　decision-making　methods　like　TOPSIS　and　SPOTIS　were　used　to　select　the　best-balanced　solution.　The　results　confirm　the　effectiveness　and　scalability　of　the　proposed　hybrid　approach,　offering　a　sustainable　pathway　to　optimize　both　economic　and　environmental　performance　in　gold　mining　operations.　The　methodology　offers　a　good　framework　for　green　mining　processes,　with　ongoing　research　aimed　at　incorporating　real-time　responsiveness　and　additional　environmental　and　economic　variables　to　further　improve　process　efficiency.　©　2025　Elsevier　Ltd