Improving the Nested Markov Chain Hyper-Heuristic (NMHH) Framework Efficiency Through Sequential Probability Ratio Testing

This paper introduces sequential probability ratio testing to the hyper-heuristic design process. Hyper-heuristics often require many function evaluations to find an acceptable heuristic configuration for the given problem. The reduction of the number of function evaluations is proposed via sequential T-testing. The new approach is compared to two state-of-the-art hyper-heuristics and six metaheuristics on twelve benchmark problems in the continuous domain. The results point to the effectiveness of the new approach.