利用多变量自适应回归样条函数确定ATCS复合分派规则的缩放参数

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．

Using MARS to determine scaling parameter values for composite dispatching rule ATCS

A multivariate adaptive regression splines (MARS) based method is proposed to determine appropriate scaling parameter values for composite dispatching rule ATCS to generate good schedules, which aims atminimizing the total weighted tardiness. With the flexiable piecewise structure in high-dimensional space, the MARS based model is able to reflect the local nonlinear relationship between the scaling parameter values and scheduling problem instances. Computational result shows that the proposed method outperforms the existing method in the literature in terms of scheduling result on total weighted tardiness and its variation.