Result files and code for a study of benchmark design, baseline choice and budget counting in RCPSP evolutionary algorithms

Result files and code behind a study of how benchmark design, baseline choice and the counting of search budgets change the conclusions drawn about a genetic algorithm and NSGA-II for resource-constrained project scheduling (RCPSP). The record contains the per-instance results for all 2,040 PSPLIB instances (five seeds each), ablations of NSGA-II, five single-pass priority rules, the genetic algorithm with forward-backward improvement on PSPLIB subsets, results for 47 empirical projects from the DSLIB database, the experiment and analysis code, tests, and a script that rebuilds the tables, figures and statistics. The PSPLIB instances and the DSLIB project files are public and are not redistributed here (see README.md).