Abstract: In order to efficiently predict the mechanical properties of ternary solid waste geopolymer grouting materials, this study conducted tests on the mechanical properties of ternary solid waste geopolymer grouting materials with different mixing ratios, and utilized the back-propagation neural network (BPNN) model and optimized it using the particle swarm algorithm (PSO) in conjunction with the SHAP (Shapley Additive exPlanations) method for an interpretable interpretability analysis. The results showed that the slag content was significantly positively correlated with the compressive strength, while the red mud content was negatively correlated with the compressive strength, the fly ash had less effect, and the exciter concentration had the most significant effect at the age of 28 days. The PSO-BPNN model outperformed the BPNN, and the coefficient of determination (R2) was improved by 0.75%.The SHAP analysis revealed that the age of maintenance and the concentration of the exciter were the main positive fac-tors affecting the compressive strength, and the red mud content had a significant negative effect on the strength. content had a significant negative effect on strength. On the untrained dataset, PSO-BPNN outperforms BPNN in terms of error fluctuation and prediction accuracy. Therefore, PSO-BPNN can provide accurate prediction of (GG) in terms of mechanical properties and guide its proportion design, which is of great significance for engineering practice.