Search

FAU - Fundação de Apoio Universitário A estimativa de produtividade é um parâmetro agronômico importante para auxiliar o mercado nacional e internacional no que tange a demanda, capacidade de transporte e armazenamento de produtos agrícolas. Nesse sentido, para que a produtividade continue a ser monitorada, tornase necessário a criação de métodos que aprimorem o mapeamento dessa variável em campo. Nesse contexto, esta pesquisa apresenta como objetivo a definição do estádio fenológico ideal e geração de modelos para estimativa de produtividade em escala local, ou seja, utilizando dados de áreas de estudo distintas para treinamento e validação do modelo. Para isso, foram utilizadas imagens multiespectrais, advindas de diferentes sensores orbitais, juntamente com índices de vegetação derivados das bandas de cada satélite, e modelos gerados por meio de algoritmos de aprendizado de máquina, sendo eles, Máquinas de Vetores de Suporte (Support Vector Machine - SVM), Redes Neurais (Neural Net - NN), Florestas Aleatórias (Random Forest - RF) e Árvores Aleatórias (Random Trees - RT). No Capítulo 1, para a definição do estádio fenológico ideal, foram utilizadas imagens do satélite Planet. Já no Capítulo 2, para a geração de modelos de estimativa de produtividade, foi utilizada uma imagem do satélite Sentinel 2. Foram utilizadas como métricas avaliativas a raiz do erro médio quadrático em percentagem (Root Mean Square Error percentagem - RMSE%) e o Erro Absoluto Médio em porcentagem (Mean Absolute Percentagem Error - MAPE) para avaliar a acurácia e tendência das estimativas de produtividade. Pela análise da definição de um estádio fenológico ideal, concluiu-se que o modelo espectral baseado em imagem da fase fenológica reprodutiva R2 foi o que obteve o melhor RMSE% de 9,17% e o segundo melhor MAPE de 7,07%. Pela análise da geração de modelos de estimativa em escala local foi possível estimar a produtividade com um RMSE% e MAPE de 20,97% e 19,19%, respectivamente. Estimated yield is an important agronomic parameter for the domestic and international market in the demand, transport capacity and storage of agricultural products. In this respect, continuous yield monitoring requires the creation of methods to enhance mapping of this variable in the field. As such, this study aimed to determine the ideal phenological stage to estimate yield and generate models for this purpose on a local scale, that is, using data from different study areas to train and validate the model. To that end, we used multispectral images from different orbital sensors combined with vegetation indices from the reflectance bands of each satellite and models generated by the following machine learning algorithms: support vector machine (SVM), neural network (NN), random forest (RF) and random trees (RT). In Chapter 1, Planet satellite images were used to determine the ideal phenological stage and in Chapter 2, yield estimate models were generated via a Sentinel 2 satellite image. The root mean square error percentage (RMSE%) and mean absolute percentage error (MAPE) were used as evaluation metrics to assess the accuracy and trend of the yield estimates. Based on analysis to determine the ideal phenological stage, it was concluded that the spectral model based on the image of phenological stage R2 obtained the best RMSE% (9.17%) and second best MAPE (7.07%). The local-scale estimation models generated made it possible to estimate yield with an RMSE% and MAPE of 20.97 and 19.19%, respectively. Dissertação (Mestrado)

Submitted by Sheila MONTEIRO (sheila.monteiro@ufma.br) on 2021-11-13T21:40:38Z No. of bitstreams: 1 MARCOS-FERREIRA.pdf: 372455 bytes, checksum: 3b2fc1435bf1e79c05a9ee6a6684f0ce (MD5) Made available in DSpace on 2021-11-13T21:40:38Z (GMT). No. of bitstreams: 1 MARCOS-FERREIRA.pdf: 372455 bytes, checksum: 3b2fc1435bf1e79c05a9ee6a6684f0ce (MD5) Previous issue date: 2021-06-29 Fundação de Amparo à Pesquisa e ao Desenvolvimento Científico e Tecnológico do Maranhão - FAPEMA Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES Acute myeloid leukemia (AML) is part of a group of hematologic neoplasms that are characterized by maturation blockade and exacerbated clonal proliferation of myeloid progenitors in the bone marrow or peripheral blood. Currently, the main diagnostic methods for this neoplasm depend on the presence of at least 20% of malignant cells in the bone marrow or peripheral blood, which generally coincides with the advanced stage of the disease. In addition, these methodologies require expensive equipment and supplies, as well as specialized labor, which makes the diagnosis of AML restricted only to large diagnostic or research centers. Therefore, in this work, we implement the use of Fourier Transform Infrared Spectroscopy and attenuated total reflection (ATR-FTIR) univariate and coupled with artificial intelligence algorithms, as a fast, inexpensive tool with high sensitivity and specificity for patient discrimination with clinical and laboratory diagnosis of acute myeloid leukemia. For this, we performed a quantitative spectral analysis of plasma (1μl) of 35 patients and 35 control subjects, with a resolution of 4 cm- 1 and 32 spectral scans in triplicate, using the air spectrum as the background of the analysis. As a result, ten vibrational modes were identified (3095 cm-1, 1648 cm-1, 1635 cm-1, 1540 cm-1, 1284 cm-1, 1170 cm-1, 1120 cm-1, 1078 cm-1, 1031 cm-1 and 850 cm-1) with sensitivity and specificity ≥80% (p

Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Marketing Intelligence This study aims to contribute to a better understanding of the problematic of data quality problems. In these days, information took an important role in every organization, a data is a valuable asset, and we discuss the types of errors in this matter and look to the capabilities of new Artificial Intelligence tools to deal with those kinds of problems. Several common problems of data quality will be identified, analysed and paired with relevant Artificial Intelligence tools resulting in a conceptual framework that information managers and data officers can use to improve that problems that we identified as critical in this current scenario. Although, it will be not possible to present an exhaustive list of all existing solutions, due to the quick emerging of new techniques and ideas about this problems