Uma busca pelo desenho amostral ótimo no problema de redução de malha amostral utilizando algoritmos genéticos: aplicado ao sistema de ovitrampas da cidade do Rio de Janeiro

• Ano de defesa: 2016
• Autor(a) principal: Costa, Leonardo Rodrigues Mattos da
• Orientador(a): Coelho, Flávio Codeço, Ferreira, Gustavo da Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Algoritmos genéticos; Otimização combinatória; Redução de malha amostral
• Palavras-chave em Inglês: Sampling grid reduction; Genetic algorithms; Qcombinatorial optimization
• Link de acesso: http://hdl.handle.net/10438/16982

Concerns about the spread of disease affect both the population in their daily life and public health policy in Brazil and worldwide. The Brazilian Ministry of Health estimates that 2.5 billion people across the globe live in regions where dengue fever is an endemic disease, and approximately 50 million people are infected each year. In Brazil, new disease outbreaks occur every 3 to 5 years, generally associated with the introduction of a new serotype in the country. The last outbreak occurred in 2013 with the introduction of the so-called type-4 dengue virus. Surveillance of mosquito reproduction and infestation includes, among other methods, the use of "ovitraps" – traps where mosquitoes lay their eggs, which are considered one the best alternatives for detecting dengue and yellow fever outbreaks. The aim of this study is to reduce the sample size of the system for capturing dengue mosquito eggs used by the city of Rio de Janeiro. With this, it will be possible to increase the frequency of data collection from a monthly to a weekly basis without increasing costs, while maintaining the quality of estimates obtained from the sample. The sampling grid reduction problem is associated with a combinatorial optimization problem belonging to the NP class where, given a sample of size n, a subset of elements of size n* needs to be found such that the estimation error is less a preset limit. From this definition, it is possible to turn the sample reduction problem into a case of the 0/1 knapsack problem. Using this association, this paper proposes objective functions that incorporate spatio-temporal dependence effects as well as an approach using biased random-key genetic algorithms