Abstract: To address the problem of high filling cost in an open pit to an underground mine, based on the machine learning method, the filling cementitious material needed for subsequent backfill mining method was developed using the available industrial wastes around the mine, and the ratio of filling slurry was optimized. First, the physical and chemical properties of the materials were analyzed. Unconfined compressive strength tests were conducted with different activator formulations to analyze the influence of each component on the strength of the backfill body. A genetic algorithm and support vector machine (GA?SVM) model was established to predict the steel-slag-based cementitious material formula using the experimental data, and the optimal ratio was determined based on the model prediction results. X-ray diffraction (XRD) and scanning electron microscope (SEM) were used to analyze the hydration products and microstructure characteristics of steel-slag-based cementitious materials at different curing ages and slag dosage conditions and determine the hydration mechanism of steel-slag-based cementitious materials. Finally, the slurry proportion was optimized by strength (i.e., 7 and 28 days) and working characteristics (i.e., slump and bleeding rate) based on the principle of gray target decision. Results revealed that the relative errors of the GA?SVM model for predicting the steel-slag-based cementitious materials strength at 7 and 28 days are 3.6%–12.62% and 6.9%–10.19%, respectively, thereby indicating high prediction accuracy. The optimal proportion of steel-slag-based cementitious materials determined by prediction analysis is steel slag content of 30%, desulfurized gypsum content of 4%, cement clinker content of 12%, and mirabilite content of 1%. The main hydration products of steel-slag-based cementitious materials are amorphous C?S?H gel, ettringite, tricalcium aluminate hydrate, Ca(OH)₂, and CaCO₃. The calcium hydroxide content increases with the steel slag content, which generates a large number of pores and deteriorates the structure and strength of the sample. When the new steel-slag-based cementitious material is applied to the actual backfilling of the mine, the optimal ratio parameters of filling slurry are obtained through the optimization of the model of the gray target decision (i.e., cement?sand ratio of 1∶4 and mass concentration of 72%). Corresponding verification experiments were conducted, and the corresponding strength and working characteristic parameters were 1.74 MPa, 3.61 MPa, 24.2 cm, and 5.91%, which all met the requirements of subsequent filling. With this proportion, the filling cost is 113 ￥·m^?3, which is 38.92% lower than that of the cement filler. The research results will benefit the comprehensive utilization of solid waste and provide support for safe, clean, and efficient mining.