| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6.33 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Com a expansão da agropecuária brasileira, o aumento da exportação e do rigor na garantia de segurança alimentar é cada vez mais urgente existindo a necessidade de investir em tecnologias voltadas para a gestão de seus processos produtivos. Medir, planear, melhorar os resultados e indicadores zootécnicos tornam-se primordiais para aumentar a produtividade, qualidade, lucro, reduzir perdas e custos. É desta forma que o Data Mining pode apresentar novos mecanismos para gerar previsões que auxilie os criadores a agirem de forma preventiva mitigando problemas como propagação de pragas e infeções entre a produção. Assim, esta dissertação versa de um estudo de caso numa fazenda extensiva de gado bovino de corte no interior de Minas Gerais, Brasil. A metodologia do Cross Industry Standard Process for Data Mining (CRIPS-DM) foi utilizada para conduzir o desenvolvimento de um modelo de gestão introduzido em uma aplicação web e mobile com recursos a algoritmo do Machine Learning capaz de armazenar dados, disponibilizar informações, gerenciar e estruturar os processos. O sistema é focado em auxiliar os pequenos e médios pecuaristas a garantir a saúde e bem-estar do gado. A fazenda de caso de estudo está a enfrentar uma alta taxa de mortalidade, sendo um cenário muito comum entre os produtores brasileiros, pois, foi efetuada uma pesquisa sobre o atual panorama da agropecuária regional e identificou-se um alto índice de mortes devido a questões sanitárias. Os fazendeiros em questão não utilizam banco de dados nem possuem sistemas de gestão estruturados. Propõe-se então, um modelo e uma aplicação de fácil implementação, utilização, com funcionalidades customizadas à real necessidade do produtor para ajudá-los na longevidade animal, redução de doenças e crescimento da rentabilidade. A aplicação web foi totalmente desenvolvida, no entanto, o algoritmo não foi finalizado devido a limitação do curto prazo de entrega e a falta de registos históricos da fazenda, pois esta não possui a rotina de recolha e armazenamento de dados bem definida. Apresentações de demos foram realizadas com produtores da região que testaram a ferramenta e modelo de gestão avaliando-os por meio de um questionário qualitativo, onde emitiram suas opiniões sobre a usabilidade, facilidade de utilização, de compreensão e satisfação com as funcionalidades. O projeto obteve uma avaliação positiva em todos os aspectos, porém, as respostas demonstraram que eles não utilizam tecnologias de informação (TI), principalmente, devido a falta de personalização, alta complexidade e dificuldade de acesso as soluções disponíveis no mercado. De acordo com os resultados da pesquisa o custo-benefício não é um fator determinante na não adoção de TI.
Brazilian agriculture is in expansion, also the exportation and the rigor in food safety. Because of that, they need to invest in technologies aimed at the management of its production processes is increasingly urgent. Measuring, planning, improving results, and zootechnical indicators have become essential to increase productivity, quality, profit, reduce losses and costs. This is how Data Mining can present new mechanisms to generate forecasts that help breeders to act preventively, mitigating problems such as the spread of pests and infections between the production. Thus, this dissertation is about a case study in an extensive beef cattle farm in Minas Gerais, Brazil. The CRIPSDM methodology leads the development of a management model and a web and mobile application with Machine Learning algorithm resources. They can store data, providing information, managing, and structuring processes. In addition, it is focused on helping small and medium-sized ranchers to ensure the health and well-being of their cattle. The case study farm is facing a high mortality rate, which is a very common scenario among Brazilian producers. According to research carried out, the farms are having a high rate of deaths due to sanitary issues. The farm does not use a database and does not have a management systems structure. It is then proposed a model and an application that is easy to implement, use, with features customized to the real needs of the producer to help them in animal longevity, disease reduction, and profitability growth. The application was developed, however, the algorithm was not finalized due to the limitation of the short delivery time and the lack of historical records of the farm, as it does not have a well-defined data recovery and storage routine. The producers could test and evaluate them through a questionnaire. They evaluated the usefulness, how is easy to use, and to understanding and satisfaction of the application and management model. The project obtained a positive evaluation in all aspects. However, the answers showed that they do not use information systems, mainly due to lack of customization, high complexity, and difficulty in accessing solutions available on the market. According to the survey results, cost-effectiveness is not a determining factor in the nonadoption of systems.
Brazilian agriculture is in expansion, also the exportation and the rigor in food safety. Because of that, they need to invest in technologies aimed at the management of its production processes is increasingly urgent. Measuring, planning, improving results, and zootechnical indicators have become essential to increase productivity, quality, profit, reduce losses and costs. This is how Data Mining can present new mechanisms to generate forecasts that help breeders to act preventively, mitigating problems such as the spread of pests and infections between the production. Thus, this dissertation is about a case study in an extensive beef cattle farm in Minas Gerais, Brazil. The CRIPSDM methodology leads the development of a management model and a web and mobile application with Machine Learning algorithm resources. They can store data, providing information, managing, and structuring processes. In addition, it is focused on helping small and medium-sized ranchers to ensure the health and well-being of their cattle. The case study farm is facing a high mortality rate, which is a very common scenario among Brazilian producers. According to research carried out, the farms are having a high rate of deaths due to sanitary issues. The farm does not use a database and does not have a management systems structure. It is then proposed a model and an application that is easy to implement, use, with features customized to the real needs of the producer to help them in animal longevity, disease reduction, and profitability growth. The application was developed, however, the algorithm was not finalized due to the limitation of the short delivery time and the lack of historical records of the farm, as it does not have a well-defined data recovery and storage routine. The producers could test and evaluate them through a questionnaire. They evaluated the usefulness, how is easy to use, and to understanding and satisfaction of the application and management model. The project obtained a positive evaluation in all aspects. However, the answers showed that they do not use information systems, mainly due to lack of customization, high complexity, and difficulty in accessing solutions available on the market. According to the survey results, cost-effectiveness is not a determining factor in the nonadoption of systems.
Description
Keywords
Pecuária Aplicação de gestão Modelo de gestão Naive Bayes Livestock Management application Management model