To address the persistent challenge of accurately predicting the unconfined compressive strength(UCS)of cemented paste backfill(CPB)in underground mining,this study developed a hybrid intelligent model integrating quantum-behaved particle swarm optimization(QPSO)with a fully connected neural network(FCNN).A comprehensive database comprising 383 laboratory test samples was compiled,encompassing eight input variables:coefficient of uniformity(Cu),chemical composition(CaO and MgO),cement type(CT),tailings-to-cement(T/C)ratio,slurry mass concentration(MC),curing temperature(Temp),and curing time(Time).The QPSO algorithm was employed to perform a global optimization of critical hyperparameters,including hidden nodes,learning rate,regularization coefficient,dropout rate,and batch size.Subsequently,the Adam optimizer was applied for deterministic fine-tuning of network weights,yielding concurrent enhancements in convergence efficiency and predictive accuracy.Furthermore,SHAP,±10%perturbation analysis,and partial dependence plots(PDPs)were utilized to quantitatively evaluate the global importance,local marginal effects,and interactive sensitivities of input parameters.The proposed QPSO-FCNN model exhibited outstanding predictive performance,achieving an R2of 0.969,RMSE of 0.198 MPa,mean absolute erron(MAE)of 0.139 MPa,and a composite performance score(CPS)of 0.930,surpassing both conventional FCNN and QPSO-BPNN hybrid models.Sensitivity analysis revealed that T/C and Time were the most influential factors,while pronounced coupling effects occurred between Cu and MC,as well as between curing conditions and T/C.These results provide robust quantitative insights for optimizing the mix design and curing regimes of CPB,thereby enhancing its mechanical reliability in underground mining operations.