Abstract: Grouting is the main treatment method of mining subsidence area, and the grouting materials have great influence on the grouting effect and cost. Taking the grout mix composed of loess and cement used in northern Shaanxi as the research object, the viscosity, the gelation time, the stone rate and 28 d uniaxial unconfined compressive strength of loess grout were studied by single factor and orthogonal tests. The influencing factors and changing rule of each index were analyzed. Meanwhile, the GA-BPNN prediction model of loess-based grout was established, and the multi-objective global optimization was carried out through GA. The grout ratio with the lowest cost and certain strength was optimized. The test results show that the loess-cement grout was similar to the grout of cement and fly ash commonly used in northern Shaanxi. The cost was greatly reduced, and the density and stone ratio of loess-based grouting materials are most affected by the water-solid ratio. The viscosity is mainly affected by the water-solid ratio and the loess(cement) content at the same time, the 28d strength is mainly affected by the loess(cement) content, and by the model constructed with GA-BPNN, and the relative error between the predicted value and expected value of performance of loess-based grouting materials is 0.74%-1.83%. The predicted result has high precision and could meet the requirements of practical application.