Hyperheuristics for managing a large collection of low level heuristics to schedule personnel
Peter Cowling, Konstantin Chakhlevitch
Abstract
Peter Cowling, Konstantin Chakhlevitch
Abstract
We investigate the performance of several hyperheuristics applied to a real-world personnel-scheduling problem. A hyperheuristic is a high-level search method which manages the choice of low level heuristics, making it a robust and easy to implement approach for complex real-world problems. We need only to develop new low level heuristics and objective functions to apply a hyperheuristic to an entirely new problem. Although hyperheuristic methods require limited problem-specific information, their performance for a particular problem is determined to a great extent by the quality of low level heuristics used. We address the question of designing the set of low level heuristics for the problem under consideration. We construct a large set of low level heuristics by using a technique which allows us to "multiply" partial low level heuristics. We apply hyperheuristic methods to a trainer scheduling problem using commercial data from a large financial institution. The results of the experiments show that simple hyperheuristic approaches can successfully tackle a complex real-world problem provided that low level heuristics are carefully selected to treat various constraints. We also examine how the choice of different sets of low level heuristics affects solution quality.
OpenAlex reports 59 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We investigate the performance of several hyperheuristics applied to a real-world personnel-scheduling problem. A hyperheuristic is a high-level search method which manages the choice of low level heuristics, making it a robust and easy to implement approach for complex real-world problems. We need only to develop new low level heuristics and objective functions to apply a hyperheuristic to an entirely new problem. Although hyperheuristic methods require limited problem-specific information, their performance for a particular problem is determined to a great extent by the quality of low level heuristics used. We address the question of designing the set of low level heuristics for the problem under consideration. We construct a large set of low level heuristics by using a technique which allows us to "multiply" partial low level heuristics. We apply hyperheuristic methods to a trainer scheduling problem using commercial data from a large financial institution. The results of the experiments show that simple hyperheuristic approaches can successfully tackle a complex real-world problem provided that low level heuristics are carefully selected to treat various constraints. We also examine how the choice of different sets of low level heuristics affects solution quality.
Key concepts: Heuristics, Hyper-heuristic, Computer science, Scheduling (production processes), Set (abstract data type), Job shop scheduling, Mathematical optimization, Schedule