In order to predict the filling mud strength and fluidity of highway subgrade with rockfill method quickly and accurately,the least square fitting is introduced into traditional GEP algorithm, and the LS-GEP efficient function mining model is established. Based on the analysis of the key characteristics and influencing factors of the large particle size filling mud, three key parameters of lime soil ratio, water soil ratio and plastic-solid ratio are put forward. Through 30 groups of mud performance tests with different mix ratios,learning samples were obtained,and LS-GEP efficient function mining model was used to predict the strength and fluidity of the filling mud , The results indicate that the two prediction formulas are accurate and reliable, and can provide effective guidance for the design of the mix ratio of the filling mud. In the actual engineering of the Hebei section of Yanqing to Chongli Expressway, the intelligent prediction formula of mud characteristics is applied to determine the mud mix quickly. During the construction process, various performance indicators of the filling mud have reached the expectations. The abutment subgrade constructed by using large particle size rockfill roadbed filling technology presents excellent performance, and no significant settlement has been observed, which the problem of vehicle jumping at the abutment has been solved effectively.