Gamification of Optimisation for Operations Research
Guardado en:
| Publicado en: | PQDT - Global (2024) |
|---|---|
| Autor principal: | |
| Publicado: |
ProQuest Dissertations & Theses
|
| Materias: | |
| Acceso en línea: | Citation/Abstract Full Text - PDF Full text outside of ProQuest |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Resumen: | Many operational research problems such as the 2D Bin-Packing Problem (2DBPP) are categorised as NP-Hard. For these problems it is either very unlikely or simply not feasible to find the global optimum within a reasonable time limit, so instead optimisation techniques are applied to find the best solution possible within a given period or number of function evaluations. These techniques include mathematical approaches, heuristics, algorithms, and metaheuristics, and techniques that combine two or more of these approaches such as hyper-heuristics. These vary from those which are highly specialised on a single problem to those which are generalised enough to optimise any problem.Humans are capable of optimising and solving real world equivalents of operational research problems, often creating and employing heuristics to do so. While many of these heuristics are consciously designed by the creator, others will be instinctive or intuitive and rely on a subconscious understanding of the problem that the creator might struggle to put into words. These latter heuristics are difficult to define or capture.Humans still struggle as problems increase in scale or complexity, but gamification includes a range of techniques that assist with many human-computer interaction scenarios. Gamification techniques can be used to make complex things more easily understandable, to help focus or retain user attention, and to encourage user engagement. This research created a gamified version of a simplified operational research problem and recorded the actions of participants interacting with that game.Optimisation methods can be either deterministic, stochastic, or a combination of the two. Deterministic approaches favour exploitation of the search space while stochastic approaches favour exploration; a combination of the two results in a method that is good at both exploration and exploitation. This research derived deterministic human heuristics from data obtained through participants playing the gamified version of an operational research problem and combined it with a generalist stochastic metaheuristic to produce a hybrid approach. This took the form of a hybrid genetic algorithm (GA).Several human-derived heuristics were compared against each other and existing heuristics such as First Fit and Best Fit on a range of test problems of increasing size and complexity. While the human-derived heuristics were generally outperformed by the existing heuristics, they outperformed the GA using the purely stochastic random mutation. A combination of stochastic and deterministic approaches was found to be more effective than either approach by itself in all cases. There is potential to apply the techniques described here to problems with no or very few known heuristics to derive new effective heuristics from humans playing a game of a simplified version of that problem. |
|---|---|
| ISBN: | 9798304988568 |
| Fuente: | Publicly Available Content Database |