A Methodology to Determine the Subset of Heuristics for Hyperheuristics through Metalearning for Solving Graph Coloring and Capacitated Vehicle Routing Problems

In this work, we focus on the problem of selecting low-level heuristics in a hyperheuristic approach with offline learning, for solution instances different domains. The objective is to improve performance approach, identifying equivalence classes set problems and best performing each them. A methodology proposed as first step all problems, generic characteristics instance one them are considered define vectors make grouping classes. Metalearning statistical tests used select class. Finally, Naive Bayes test k-fold cross-validation, compared results statistically best-known values. research, was tested by applying it capacitated vehicle routing (CVRP) graph coloring (GCP). experimental show that can correctly appropriate case. This based comparison obtained those state art instance.