<p>Accurate prediction of tribological behavior in aluminum matrix composites (AMCs) remains a significant challenge due to complex nonlinear interactions between operational parameters and material properties. This study develops and validates a machine learning (ML)-based predictive framework for dry sliding wear rate estimation across multiple AMC systems. An experimental tribological database comprising 156 pin-on-disc wear tests was compiled, covering diverse aluminum alloy matrices reinforced with ceramic particles such as SiC, B₄C, Al₂O₃, and TiC, subjected to varying normal loads, sliding speeds, and sliding distances. Seven ML algorithms Gradient Boosting (GB), Random Forest (RF), AdaBoost, Gaussian Process Regression (GPR), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Linear Regression (LR) were systematically trained, validated, and compared using R², Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) as performance indicators. Gradient Boosting emerged as the superior model, achieving R² test values of 0.965 and 0.986 alongside MAE values of 0.082 and 0.072 and RMSE values of 0.115 and 0.098 for SiC- and B₄C-reinforced composites, respectively, confirming its strong predictive capability through robust statistical evidence. Microstructural examination of worn surfaces revealed wear mechanism transitions from adhesive and oxidative modes at moderate speeds to abrasive-dominated behavior at elevated loads, consistent with model predictions. The findings demonstrate that ensemble learning methods, particularly Gradient Boosting, offer an effective and computationally efficient approach for accelerating AMC design by reducing experimental characterization effort.</p>