In legacy software, non-functional concerns tend to cut across the system and manifest themselves as tangled or scattered code. If these crosscutting concerns could be modularized and the system refactored, then the system would become easier to understand, modify, and maintain. Modularized crosscutting concerns are known as aspects and the process of identifying aspect candidates in legacy software is called aspect mining. One of the techniques used in aspect mining is clustering and there are many clustering algorithms. Current aspect mining clustering algorithms attempt to form clusters by optimizing one objective function. However, the objective function to be optimized tends to bias the formation of clusters towards the data model implicitly defined by that function. One solution is to use algorithms that try to optimize more than one objective function. These multiobjective algorithms have been used successfully in data mining but, as far as this author knows, have not been applied to aspect mining. This study investigated the feasibility of using multiobjective evolutionary algorithms, in particular, multiobjective genetic algorithms, in aspect mining. The study utilized an existing multiobjective genetic algorithm, MOCK, which had already been tested against several popular single objective clustering algorithms. MOCK has been shown to be, on average, as good as, and sometimes better than, those algorithms. Since some of those data mining algorithms have counterparts in aspect mining, it was reasonable to assume that MOCK would perform at least as good in an aspect mining context. Since MOCK's objective functions were not directly trying to optimize aspect mining metrics, the study also implemented another multiobjective genetic algorithm, AMMOC, based on MOCK but tailored to optimize those metrics. The reasoning hinged on the fact that, since the goal was to determine if a clustering method resulted in optimizing these quality metrics, it made sense to attempt to optimize these functions directly instead of a posteriori. This study determined that these multiobjective algorithms performed at least as good as two popular aspect mining algorithms, k-means and hierarchical agglomerative. As a result, this study has contributed to both the theoretical body of knowledge in the field of aspect mining as well as provide a practical tool for the field.

[pt] A programação de produção em refinaria pode ser compreendida como decisões que buscam otimizar alocação de recursos, o sequenciamento de atividades e a sua realização temporal, respeitando restrições e visando ao atendimento de múltiplos objetivos. Apesar da complexidade e natureza combinatória, a atividade carece de sistemas sofisticados que auxiliem o processo decisório, especialmente baseadas em otimização, pois as ferramentas utilizadas são planilhas ou softwares de simulação. A diversidade de objetivos do problema não implica em equivalência de importância. Pode-se considerar que existem grupos, onde os que afetam diretamente a capacidade produtiva da refinaria se sobrepõem aos associados à maior continuidade operacional. Esta tese propõe o desenvolvimento de algoritmos multiobjetivos para programação de petróleo em refinaria. As propostas se baseiam em conceituadas técnicas da literatura multiobjetivo, como dominância de Pareto e decomposição do problema, integradas à programação genética com inspiração quântica. São estudados modelos em um ou dois níveis de decisão. A diferenciação dos grupos de objetivos é avaliada com base em critérios estabelecidos para considerar uma solução proposta como aceitável e também é avaliada a influência de uma população externa no processo evolutivo. Os modelos são testados em cenários de uma refinaria real e os resultados são comparados com um modelo que trata os objetivos de forma hierarquizada. As abordagens baseadas em dominância e em decomposição apresentam vantagem sobre o algoritmo hierarquizado, e a decomposição é superior. Numa comparação com o modelo em dois níveis de decisão, apenas o que utiliza estratégia de decomposição em cada nível apresenta bons resultados. Ao final deste trabalho é obtido mais de um modelo multiobjetivo capaz de oferecer um conjunto de soluções que atendam aos objetivos críticos e deem flexibilidade de análise a posteriori para o programador de produção, o que, por exemplo, permite que ele pondere questões não mapeadas no modelo. / [en] Refinery scheduling can be understood as a set of decisions which aims to optimize resource allocation, task sequencing, and their time-related execution, respecting constraints and targeting multiple objectives. Despite its complexity and combinatorial nature, the refinery scheduling lacks more sophisticated support decision tools. The main systems in the area are worksheets and, sometimes, simulation software. The multiple objectives do not mean they have the same importance. Actually, they can be grouped whereas the objectives related to the refinery production capacity are more important than the ones related to a smooth operation. This thesis proposes the development of multiobjective algorithms applied to crude oil refinery scheduling. The proposals are based on the major technics of multiobjective literature, like Pareto dominance and problem decomposition, integrated with a quantum-inspired genetic programming approach. One and two decision level models are studied. The difference between groups is handled with conditions that define what can be considered a good solution. The effect of using an archive population in the evolutionary process is also evaluated. The results of the proposed models are compared with another model that handles the objectives in a hierarchical logical. Both decomposition and dominance approaches have better results than the hierarchical model. The decomposition model is even better. The bilevel decomposition method is the only one, among two decision levels models, which have shown good performance. In the end, this work achieves more than one multiobjective model able to offer a set of solutions which comprises the critical objectives and can give flexibility to the production scheduler does his analysis. Therefore, he can consider aspects not included in the model, like the forecast of crude oil batches not scheduled yet.