The assessment of Mean Annual Increment (MAI) is crucial for the sustainable management of forest resources. This study aims to develop predictive models of MAI in volume for Atlas cedar (Cedrus atlantica) in the Ain Leuh forest of Morocco's Middle Atlas region using advanced machine learning (ML) techniques. An extensive dataset consisting of 1,921 sample plots obtained from Morocco's National Forest Inventory, including variables such as stand density, total volume, basal area, and MAI. Additional predictor variables included topographic, bioclimatic, soil, remote sensing, and stand structure data. Pearson's correlation was used to explore relationships between MAI and other dendrometric parameters, while Random Forest (RF) and XGBoost algorithms were employed to predict MAI. The results revealed strong correlations between MAI and other dendrometric variables, indicating that MAI can serve as a proxy for modeling other forest attributes. MAI in volume was found to be influenced by climatic, edaphic, and stand-related factors. Although regression performance was limited, both models showed moderate performance in classification tasks using quantile-based MAI classes, with AUC values of 0.639 for RF and 0.641 for XGBoost. Spatial predictions indicated that the highest MAI values are concentrated in the central part of the forest, corresponding to its most fertile and productive sites. These findings, despite the moderate model performance, highlight the potential of ML techniques to model the MAI of C. atlantica in complex ecological settings and offer promising perspectives for developing decision-support tools in forestry, optimizing harvest planning and supporting sustainable forest management, particularly in the context of climate change.