Dissertations / Theses on the topic 'Local search approaches'

Afin d’améliorer la qualité de prédiction des techniques de classification automatique et de fouilles de données, plusieurs modèles ont été proposés dans la littérature en vue d’extraire des connaissances à partir des données. Toutefois, avec l’expansion des systèmes d’information et des technologies associées, ces techniques d’apprentissage s’avèrent de moins en moins adaptées aux nouvelles tailles et dimensions des données. On s’intéresse dans cette étude aux problèmes de grande dimensionnalité et à l’amélioration du processus d’apprentissage des méthodes de classification à travers les techniques de filtrage et de sélection d’attributs. Le problème « d’identification d’attributs pertinents » (Feature Selection Problem), tel qu’il est défini dans la littérature, relève d’une nature combinatoire. Dans le cadre de cette thèse, on s’est intéressé au développement de nouvelles techniques d’optimisation approchées et spécifiques au problème traité ainsi qu’à l’amélioration d’algorithmes existants. La conception, l’implémentation et l’étude empirique ont montré l’efficacité et la pertinence des métaheuristiques proposées<br>Although the expansion of storage technologies, networking systems, and information system methodologies, the capabilities of conventional data processing techniques remain limited. The need to knowledge extraction, compact representation and data analysis are highly motivated by data expansion. Nevertheless, learning from data might be a complex task, particularly when it includes noisy, redundant and information-less attributes. Feature Selection (FS) tries to select the most relevant attributes from raw data, and hence guides the construction of final classification models or decision support systems. Selected features should be representative of the underlying data and provide effective usefulness to the targeted learning paradigm (i.e. classification). In this thesis, we investigate different optimization paradigms as well as its adaptation to the requirements of the feature selection challenges, namely the problem combinatorial nature. Both theoritical and empirical aspects were studied, and confirm the effectiveness of the adopted methodology as well as the proposed metaheuristic based approaches

Étant donné leur faible impact environnemental, l’utilisation des véhicules électriques dans les activités de service a beaucoup augmenté depuis quelques années. Cependant, leur déploiement est freiné par des contraintes techniques telles qu’une autonomie limitée et de longs temps de charge des batteries. La prise en compte de ces contraintes a mené à l’apparition de nouveaux problèmes de tournées de véhicules pour lesquels, en plus d’organiser les tournées,il faut décider où et de combien charger les batteries. Dans cette thèse nous nous intéressons à ces problèmes au travers de quatre études. La première concerne le développement d’une métaheuristique en deux phases simple mais performante pour résoudre un problème particulier appelé "Green VRP”. Dans la seconde, nous nous concentrons sur la modélisation d’un aspect essentiel dans ces problèmes : le processus de chargement des batteries. Nous étudions différentes stratégies pour modéliser ce processus et montrons l’importance de considérer la nature non linéaire des fonctions de chargement. Dans la troisième étude nous proposons une recherche locale itérative pour résoudre des problèmes avec des fonctions de chargement non linéaires. Nous introduisons un voisinage dédié aux décisions de chargement basé sur un nouveau problème de chargement sur une tournée fixée. Dans la dernière étude, nous traitons un problème réel de tournées de techniciens avec des véhicules classiques et électriques. Ce problème est résolu par une métaheuristique qui décompose le problème en plusieurs sous-problèmes plus simples résolus en parallèle, puis qui assemble des parties des solutions trouvées pour construire la solution finale<br>Electric vehicles (evs) are one of the most promising technologies to reduce the greenhouse gas emissions. For this reason, the use of evs in service operations has dramatically increased in recent years. Despite their environmental benefits, evs still face technical constraints such as short autonomy and long charging times. Taking into account these constraints when planning ev operations leads to a new breed of vehicle routing problems (vrps), known as electricVrps (evrps). In addition, to the standard routing decisions, evrps are concerned with charging decisions: where and how much to charge. In this ph. D thesis, we address evrps through 4 different studies. In the first study, we tackle the green vehicle routing problem. To solve the problem, we propose a simple, yet effective, two-phase matheuristic. In the second study, we focus a key modelling aspects in evrps: the battery charging process. We study different strategies to model this process and show the importance of considering the nonlinear nature of the battery charging functions. InThe third study, we propose an iterated local search to tackle evrp with non-linear charging functions. We introduce a particular local search operator for the charging decisions based on a new fixedroute charging problem. The fourth and last study considers a real technician routing problem with conventional and electric vehicles (trp-cev). To tackle this problem, we propose a parallel matheuristic that decomposes the problem into a set of easier-to-solve subproblemsThat are solved in parallel processors. Then the approach uses parts of the solutions found to the subproblems to assemble final solution to the trp-cev

Nesta tese propõe-se um método heurístico para o problema de Projeto de Rede da Cadeia de Suprimentos (Supply Chain Network Design) considerando vários aspectos de relevância prática, tais como: fornecedores e matérias-primas, localização e operação de instalações, atribuição de Centros de Distribuição (CD), e grande número de clientes e produtos. Uma eficiente abordagem heurística de duas fases é proposta para a obtenção de soluções viáveis para os problemas, que inicialmente é modelado como um Programa Linear Inteiro Misto (PLIM) de grande escala. Na fase de construção, uma estratégia de Linear Programming Rounding é aplicada para se obter os valores iniciais para as variáveis de localização inteira do modelo. Simultaneamente, um método Multi-start foi desenvolvido para gerar soluções iniciais diversificadas para cada nova iteração da heurística de Rounding. Na segunda fase, dois procedimentos de Busca Local foram desenvolvidos no sentido de melhorar a solução fornecida pelo método de Rounding. Implementamos duas diferentes abordagens de Busca Local: remoção-inserção e troca. Uma técnica de Busca Tabu para orientar o procedimento de Busca Local para explorar os diferentes espaços de soluções foi desenvolvida. As formulações e algoritmos foram implementados na linguagem C++ utilizando ferramentas de otimização da COIN-OR. O método de solução foi experimentado em instâncias geradas aleatoriamente, com tamanhos diferentes em termos do número de parâmetros, tais como o número de produtos, zonas de clientes, CDs e fábricas considerando um sistema logístico de quatro níveis. As implementações computacionais mostram que o método de solução proposto obteve resultados satisfatórios quando comparados com a literatura. Para validar este método heurístico também foi usado em um caso realista, com base em dados de uma empresa de borracha que está reestruturando sua cadeia de suprimentos devido ao projeto de uma nova uma nova fábrica e produção de novos produtos. A abordagem heurística proposta revelou-se adequada para aplicação prática em um caso real de uma indústria multicommodity em um contexto determinístico.<br>In this thesis we propose a heuristic method for the Supply Chain Network Design (SCND) problem considering several aspects of practical relevance: suppliers and raw materials, location and operation facilities, distribution center (DC) assignments, and large numbers of customers and products. An efficient two-phase heuristic approach is proposed for obtaining feasible solutions to the problems, which is initially modeled as a large-scale Mixed Integer Linear Program (MILP). In the construction phase, a linear programming rounding strategy is applied to obtain initial values for the integer location variables in the model. Simultaneously, a Multi-start method was developed to generate diversified initial solutions from each new iteration in the rounding heuristic. In the second phase, two Local Search procedures were developed towards to improve the solution provided by the rounding method. We implemented two different Local Search approaches: removal-insertion and exchange. A Tabu Search technique was developed to guide the Local Search procedure to explore the different spaces of solutions. The formulations and algorithms were implemented in C++ code language using the optimization engine COIN-OR. The solution method was experimented in randomly generated instances, with different sizes in terms of the number of parameters, such as number of products, customer zones, DCs, and factories considering a four-echelon logistic system. The computational implementations show that the solution method proposed obtained satisfactory results when compared to the literature review. To validate this heuristic method was also used in a realistic case, based on data from a rubber company that is restructuring its supply chain due to the overture of a new factory, producing new products. The proposed heuristic approach proved appropriate to practical application in a realistic case of a multi commodity industry in a deterministic context.

No<br>This paper presents a new class of heuristics which embed an exact algorithm within the framework of a local search heuristic. This approach was inspired by related heuristics which we developed for a practical problem arising in electronics manufacture. The basic idea of this heuristic is to break the original problem into small subproblems having similar properties to the original problem. These subproblems are then solved using time intensive heuristic approaches or exact algorithms and the solution is re-embedded into the original problem. The electronics manufacturing problem where we originally used the embedded local search approach, contains the Travelling Salesman Problem (TSP) as a major subproblem. In this paper we further develop our embedded search heuristic, HyperOpt, and investigate its performance for the TSP in comparison to other local search based approaches. We introduce an interesting hybrid of HyperOpt and 3-opt for asymmetric TSPs which proves more efficient than HyperOpt or 3-opt alone. Since pure local search seldom yields solutions of high quality we also investigate the performance of the approaches in an iterated local search framework. We examine iterated approaches of Large-Step Markov Chain and Variable Neighbourhood Search type and investigate their performance when used in combination with HyperOpt. We report extensive computational results to investigate the performance of our heuristic approaches for asymmetric and Euclidean Travelling Salesman Problems. While for the symmetric TSP our approaches yield solutions of comparable quality to 2-opt heuristic, the hybrid methods proposed for asymmetric problems seem capable of compensating for the time intensive embedded heuristic by finding tours of better average quality than iterated 3-opt in many less iterations and providing the best heuristic solutions known, for some instance classes.

碩士<br>國立高雄第一科技大學<br>電腦與通訊工程系碩士班<br>105<br>Techniques for task matching and scheduling play a key role in distributing the computing resources of a heterogeneous computing environment which has become increasingly common today. We consider the task matching and scheduling with the objective of minimizing makespan. Task matching and scheduling is known to be NP-complete in the strong sense. Since this problem is extremely complex, numerous research efforts have focused on many metaheuristic approaches to obtain near-optimal solutions within a reasonable amount of computation time. In this study, we present a local search algorithm based on runtime-critical path, that can reduce the schedule length produced by any scheduling algorithm.

碩士<br>國立交通大學<br>運輸與物流管理學系<br>105<br>Multi-compartment vehicle routing problem (MCVRP) is an extension of the classical Vehicle Routing Problem (VRP) with multiple products that must be stored in the given compartments in the vehicle. The main difference between MCVRP and VRP is the compartment capacity limit for various products. Depending on if the different products of a customer can be shipped by different vehicles, the MCVRP can be classified into the split and non-split type of problems respectively. In this thesis, we apply the particle swarm optimization (PSO) method and a multi-start iterated local search (MS_ILS) approach for solving the MCVRP respectively. In the PSO part, we propose an n-dimension decoding method for solving both the non-split and split type of the MCVRP. In our MS_ILS, we first use the n-dimension decoding method to construct multiple initial solutions. Then, we apply a randomized variable neighborhood decent (RVND) module with the local search operators of 2-opt, Or-opt, λ-interchanges and 2-opt* to improve the solution. A p-point perturbation is also applied in the MS_ILS to increase the diversification of search. Our proposed algorithms were tested on four problem instance sets proposed by Fallahi et al. [9] and Reed et al. [24]. The results show that the MS_ ILS performs better than PSO. On the other hand, our proposed n-dimension decoding method of PSO can save about half computer time than the original 2n-dimension decoding method in solving the MCVRP for the split type of problems. Out of 136 instances tested, our MS_ILS has found 23 best known solutions (BKS) and 38 new best solutions. The average deviation from BKS is 0.18%. Key Words: Multi-Compartment Vehicle Routing Problem (MCVRP), Particle Swarm Optimization (PSO), Iterated Local Search (ILS), Multi-start