Análise de competição familiar por algoritmos genéticos inspirados em modelos cinéticos de mercado

• Autor(a) principal: Luquini, Evandro
• Data de Publicação: 2019
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações do Mackenzie
• Texto Completo: http://dspace.mackenzie.br/handle/10899/24297
• topic: algoritmos evolutivos; econofísica; modelos de mercado de câmbio; modelos cinéticos de mercado; modelos de desigualdade econofísicos; otimização computacional; meta-heurísticas de otimização; algoritmos genéticos de competição familiar; diversidade populacional; CNPQ::CIENCIAS EXATAS E DA TERRA

Family competition genetic algorithms are a variant of classical genetic algorithms that evolveapopulationbyonlyemphasizingthesubstitutionrules.Inthisversion,selectionfor reproductionfavoringthebestindividualsisreplacedbyastrategythatgivesequalchances to all members of the population. The individuals selected to remain in the population consist solely of those within the family, which is formed by the pair of individuals selected for reproduction and their descendants, without considering the general state of the simulation. Over time, several studies have been done to find new substitution rules to improve population diversity and the effectiveness of this approach. Our study aligns with this goal and introduces new substitution rules inspired by the kinetic market models of econophysics. These models have many similarities with family competition genetic algorithms, but, differing from them, kinetic market models were designed so that the simulation results in a stationary distribution. The dynamics of a kinetic market model resemble a minimization procedure because all the agents are progressively shifted to lower energy states, independent of the initial distribution. However, these models are capable of preventingcondensation:thesituationinwhichallagentsinthepopulationconvergetothe samestateorwhenoneagenthijacksallthesystemenergy.Thesepropertiesareusedinthis work to increase the diversity and effectiveness of family competition genetic algorithms. Through a statistical protocol, the new substitution rules were tested and analyzed against others in the literature. The results were positive for experiments with combinatorial problems. In addition, the introduction of evolutionary computing individuals into the kinetic market also produces results for econophysics. The new non-conservative kinetic market model that results from the proposed merging differs from its peers by displaying random walks for the sum of all agents’ money and simultaneously a scaling behavior for their distribution in the population