Anchor and fissure grouting are used to repair earthen sites. However, the common method to obtain the compressive strength of grouting slurry would cause material, labor, and time losses. In addition the material properties, environmental and economic benefits have gained increasing attention. This study proposes a design framework for multi-objective proportioning optimization based on machine learning and metaheuristics. The results indicated that the eXtreme Gradient Boosting (XGBoost) model, whose hyper-parameters were optimized by a genetic algorithm, can accurately predicted the compressive strength of the slurries. The impact of the variables on development of compressive strength can explain the internal reaction mechanisms. The analytical framework based on meta-heuristic and technique for order of preference by similarity to an ideal solution (TOPSIS) provided Pareto-optimal solutions in design scenario of each sub-dataset. The framework proposed in this study can efficiently achieve mechanical, environmental, and economic design objectives of anchor grouting and fissure grouting slurries for repairing earthen sites.