Fabric Reinforced Cement Mortar (FRCM) jacketing is an effective intervention to increase the shear strength of deficient structural members. However, research on predicting shear strength in reinforced concrete (RC) beams strengthened with FRCM is limited. Previous studies have used mechanical models and empirical formulas for shear capacity estimation, often falling short due to complex phenomena. To achieve more accurate predictions, civil engineering research has turned to machine learning, with boosting algorithms emerging as the preferred choice for tabular data. These algorithms iteratively adjust training data weights to enhance predictive accuracy by prioritizing previously misclassified samples. Noted for their robust performance, generalization capabilities, and reliability, boosting algorithms offer a promising approach. This study validates the superiority of boosting algorithms for predicting the shear strength of FRCM strengthened RC beams by comparing Extreme Gradient Boosting (XGBoost) with linear regression (LR) and Support Vector Machines (SVM). Assessing both tree-based and non-tree-based approaches, the research underscores the predictive accuracy of boosting algorithms. Utilizing a dataset of 174 beams with 14 input parameters, the data is partitioned into 70% training and 30% test sets. Results indicate that XGBoost outperforms other algorithms, achieving the highest determination coefficient (R2 = 0.9898 versus 0.9396 for LR and 0.7112 for SVM), the smallest mean absolute error (MAE = 5.360 kN), and the lowest root mean square error (RMSE = 10.071 kN). The SHapley Additive exPlanations (SHAP) method elucidates the key input features in XGBoost in descending order: beam depth, concrete compressive strength, with mortar thickness also notable.