Experimental dataset: Neighbourhood Structures and Ranking Operators for the Interval Job Shop Scheduling Problem

The experimental study compares five neighbourhood variants($N_1$, $N_2$, $N_3$, $N_{\text{ext}}$, $N_8$) and four interval-ranking operators(EV, LEX1, LEX2, YX) for the Interval Job Shop Scheduling Problem (IJSP), on 82benchmark instances ranging from $10\times10$ to $50\times20$ operations. Threeexperimental phases are reported: - **Phase A** — common-hyperparameter hill climbing for the $5 \times 4$ operator comparison (Section 8.2 of the paper).- **Phase B** — per-neighbourhood irace-tuned tabu search for the head-to-head neighbourhood comparison (Section 8.3).- **Phase C** — comparison of the best Phase B configuration against three published IJSP solvers (Section 8.4). Each (instance, configuration) pair is executed for 30 independent runs.本实验针对区间作业车间调度问题（Interval Job Shop Scheduling Problem, IJSP）展开，对比了5种邻域变体（$N_1$、$N_2$、$N_3$、$N_{ ext{ext}}$、$N_8$）与4种区间排序算子（EV、LEX1、LEX2、YX），实验数据集包含82个基准实例，作业规模覆盖$10 imes10$至$50 imes20$。本次研究报告了三个实验阶段： - **阶段A** — 面向$5 imes4$算子对比的通用超参数爬山算法（对应论文8.2节）。 - **阶段B** — 针对邻域两两对比的、经各邻域irace调优的禁忌搜索算法（对应论文8.3节）。 - **阶段C** — 将阶段B所得最优配置与3种已发表的IJSP求解器进行对比（对应论文8.4节）。 每组（实例，配置）组合均执行30次独立运行。