A methodology for determining an effective subset of heuristics in selection hyper-heuristics.

•A methodology for determining an effective subset of heuristics in hyper-heuristics.•Use of fitness landscape metrics and non-parametric statistics to rank heuristics.•New adaptive iterated local search hyper-heuristic using dynamic multi-armed bandit.•State-of-the-art results in course timetabling including new best-known solutions.