提出一种利用多变量自适应回归样条函数(MARS)来确定考虑准备时间的直观延误成本(ATCS)复合分派规则中缩放参数的方法,以优化ATCS分派规则在最小化总加权延误时间(TWT)上的效果.通过利用MARS模型在高维空间上的弹性建模能力,构建调度作业组与缩放参数之间的非线性模型,以便灵活地捕捉更多的局部映射关系.对比实验结果表明,与已有方法相比,该方法可显著地改善ATCS规则在最小化总加权延误时间上的效果,同时降低调度效果的不稳定性. This paper proposes a method to determine the scaling parameters in the visual delay cost (ATCS) compound assignment rule considering the preparation time by using multivariate adaptive regression spline function (MARS) to optimize the performance of the ATCS assignment rule in minimizing the total weighted delay time (TWT ) By using the flexible modeling ability of MARS model in high-dimensional space, a nonlinear model between scheduling job group and scaling parameters is constructed to capture more local mapping relationships flexibly.Comparison results show that, Compared with the existing methods, this method can significantly improve the effect of ATCS rules on minimizing the total weight delay and reduce the instability of scheduling effects.