<p>Well productivity plays a vital role in determining the economic viability of hydrocarbon field development. Among stimulation techniques, matrix acidizing is highly effective for improving well performance by removing near-wellbore formation damage. The success of an acidizing treatment depends on the precise selection of injection volume and rate to maximize damage removal while minimizing operational costs. This study introduces an integrated computational framework for optimizing matrix acidizing processes. The framework first predicts post-treatment outcomes and then determines optimal injection parameters. Its main objective is to identify the acid volume and injection rate that minimize the post-treatment skin factor, thereby enhancing well productivity while reducing acid consumption and cost. The methodology consists of two sequential phases: prediction and optimization. In the prediction phase, four machine learning algorithms—Extra-Trees (ExTree), Random Forest, Gradient Boosting, and AdaBoost—were trained using a comprehensive dataset derived from extensive reservoir simulations to estimate post-acidizing skin factor and injection pressure. Among these models, the ExTree algorithm achieved the highest predictive accuracy, with an R² value of 0.9390 for the skin factor. In the optimization phase, the validated ExTree model was coupled with two metaheuristic algorithms, Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), to determine the optimal injection parameters. Both optimization algorithms consistently converged to stable solutions within 500 iterations. The framework’s computational cost is practical for engineering design and offline analysis, confirming its feasibility for field applications. The novelty of this research lies in its unified, data-driven approach that combines high-accuracy prediction with advanced optimization techniques. This integration provides a robust and efficient tool for the rational design of well stimulation treatments.</p>