Research Area: Metaheuristic Computing
Internet Interdisciplinary Institute (IN3) Doctor of Network and Information Technologies From Metaheuristics to Learnheuristics: Applications to Logistics, Finance, and Computing by Laura Calvet Linn A large number of decision-making processes in strategic sectors such as transport, production and finance involve NP-hard problems. Trends such as globalization make systems larger and more complex. Frequently, these problems are characterized by high levels of uncertainty and dynamism. Metaheuristics have become predominant methods for solving challenging optimization problems in reasonable computing times.
From a methodological perspective, the main contributions of this thesis are the design of learnheuristics and a classification of works hybridizing statistical / machine learning and metaheuristics. It discusses the potential of learn heuristics in a number of fields and studies two specific cases. The first is a routing problem in which the depots are heterogeneous, in terms of their commercial offer, and customers show different willingness to consume depending on how well the assigned depot fits their preferences. Thus, different customer-depot assignment maps lead to different customer-expenditure levels.
In the production arena, the optimization of jobs sequences under stochasticity, considering multiple production lines and a common deadline, is discussed. Strategies to invest on risky assets are proposed and assessed. Finally, the parameter fine-tuning of metaheuristics and the effect of the number of agents and the computing time on metaheuristics performance are investigated.
Name of the Researcher: Laura Calvet Linan
Name of the Supervisor(s): Angel A.Juan
Year of Completion: 2017
University: Open University of Catalonia
Thesis Link: Home Page Url
