Aluminum alloy WAAM is a complex physical system with multi-parameter coupling, and the accurate prediction and control of its forming dimensions are affected by various process parameters. Aiming at the problems of insufficient modeling of parameter coupling effect, limited prediction accuracy and lack of model interpretability in existing prediction methods, this study proposes an interpretable data-driven model based on data augmentation strategy and ensemble learning method to achieve high-precision prediction of width and layer height in aluminum alloy forming process. First, the training dataset is augmented by data augmentation techniques to enhance the generalization ability of the model. Secondly, multiple models are trained based on the five-fold cross-validation method, and three base learners with the best performance are evaluated. Then, the SCSO algorithm is used to optimize the weight allocation of the basis learner, and a highly robust ensemble learning model is constructed. Finally, the SHAP method is used to quantify and explain the effects of process parameters on the forming process. The experimental results show that the ensemble learning model based on SCSO optimization significantly outperforms the single model and the traditional ensemble learning method in the prediction accuracy and interpretability of aluminum alloy forming dimensions (RMSE is 0.3518 and 0.0743, and MAPE is 0.0229 and 0.0364 when predicting width and layer height). This study provides a  heoretical basis for process parameter optimization and forming quality control of aluminum alloy WAAM, with good practicality and engineering application value.