The intricate interaction between essential process parameters and their impact on gold–cyanide leaching recovery presents an environmental challenge for mining companies. Although polycarbonate materials demonstrate potential as adsorbents for gold-cyanide extraction, improving the adsorption process demands robust predictive models capable of accommodating variability among various operational parameters. This work proposed the development of a machine learning model that forecasts gold–cyanide leaching removal utilizing the architecture of Automated Machine Learning (AutoML) and feature engineering approaches in MATLAB. The experimental feed data comprised flow rate (900-1000 m³/hr), pH (10-11), and polycarbonate concentration (7-10 g/l) as the independent variables mapped against cyanide concentration (200-300 ppm) and gold concentration (0.5-2 ppm) as the dependent variables. The dimensionality reduction technique enabled the elimination of redundant input characteristics that exert negligible influence on gold cyanide leaching removal during the preprocessing phase. Evaluation of the regression model, following training and hyperparameter optimization, identified the decision tree (DT) as the most effective model, achieving a coefficient of determination (R²) of 0.998, a mean squared error (MSE) of 0.025, and a root mean square error (RMSE) of 0.1581. Feature selection employing the F-test identified pH as the predominant variable during model training. The proposed prediction model offers a deployable system designed to optimize mining operations and save operational expenses. This also underlines the reliability and rapid implementation of AutoML in these domains.