Hydrometallurgical　acid　leaching　is　a　promising　route　for　copper　recovery　from　copper　slag,　yet　the　wide　compositional　variability　of　slags　and　the　multitude　of　leaching　parameters　make　the　experimental　search　for　optimal　conditions　both　costly　and　labor-intensive.　To　overcome　this　bottleneck,　we　developed　a　machine-learning　(ML)　framework　that　can　rapidly　predict　and　optimize　copper　extraction.　Four　algorithms—Partial　Least　Squares　Regression　(PLSR),　Gradient　Boosting　Regression　(GBR),　AdaBoost　Linear　Regression　and　Random　Forest　(RF)—were　systematically　compared;　RF　delivered　the　highest　predictive　accuracy.　Using　slag　chemistry　and　leaching　parameters　as　inputs　and　copper　leaching　efficiency　as　the　output,　the　RF　model　achieved　an　average　R2　of　0.93,　with　RMSE　and　MAE　as　low　as　7.504　and　5.681,　respectively.　Sensitivity　analysis　revealed　that　leaching　conditions—especially　leaching　time,　acid　concentration　and　temperature—exert　a　far　stronger　influence　on　extraction　than　slag　composition　itself.　This　ML-assisted　approach　offers　an　efficient,　data-driven　strategy　for　understanding　and　optimizing　copper-slag　hydrometallurgy.　©　2025　Elsevier　B.V.