Estudo comparativo de algoritmos de classificação para predição da glicemia sangüínea em pacientes diabéticos
Carregando...
Arquivos
Data
2012
Tipo de documento
Monografia
Título da Revista
ISSN da Revista
Título de Volume
Área do conhecimento
Ciências Exatas e da Terra
Modalidade de acesso
Acesso aberto
Editora
Autores
Claudio, Onyas D'aquino
Orientador
Pereira, Max Roberto
Coorientador
Resumo
A Diabetes Mellitus e um mal que aflige um grande numero de pessoas. Nao existe cura e o tratamento e todo baseado no controle. Dessa forma, este trabalho dispoe de informacoes sobre a doenca e acompanha o desenvolvimento de um aplicativo que auxilie no controle da alimentacao, medicacao e glicemia sanguinea de um paciente. Com metodologia cientifica, foi realizado um estudo teorico sobre a Diabetes, suas causas, consequencias, formas de tratamento disponivel, englobando varios aspectos como nutricao e farmacia. Seguido das etapas de projecao e construcao de um sistema informatizado. Por fim, discutimos tecnicas de classificacao que permitiram fazer a predicao da glicemia sanguinea em pacientes diabeticos, tais como regressao linear, redes neurais artificiais e arvores de decisao. Os resultados obtidos por meio delas podem ajudar ao diabetico a definir sua dieta, fundamental no tratamento, para evitar complicacoes. Ao final, avaliamos e comparamos as tecnicas utilizadas. A regressao conseguiu um acerto de 57% por aproximacao do valor real, enquanto a arvore de decisao e a rede neural conseguiram taxas de acerto de 65% e 74%, respectivamente, por meio da classificacao dos dados. Numeros que foram discutidos e analisados ao final deste trabalho.
Diabetes Mellitus is a disease that afflicts a large number of people. There is no cure and treatment is all based on control. So, this paper proposes with information about the disease and accompanies the development of an application that aids in the control of feeding, medication and blood glucose levels of a patient. With scientific methodology, we conducted a theoretical study on diabetes, its causes, consequences, treatments available, covering various aspects such as nutrition and pharmacy. Follow by the steps of designing and building a computerized system. Finally, we discuss techniques for classification to the prediction of blood glucose in diabetic patients, such as linear regression, artificial neural networks and decision trees. The results obtained by them may help to define a diabetic diet, basic treatment to prevent complications. To finish, we evaluate and compare the techniques used. The regression took a rate of 57% by approximation of the true value, while the decision tree and the neural network achieved success rates of 65% and 74%, respectively, by classifying the data. Numbers that have been discussed and analyzed at the end of this paper.
Diabetes Mellitus is a disease that afflicts a large number of people. There is no cure and treatment is all based on control. So, this paper proposes with information about the disease and accompanies the development of an application that aids in the control of feeding, medication and blood glucose levels of a patient. With scientific methodology, we conducted a theoretical study on diabetes, its causes, consequences, treatments available, covering various aspects such as nutrition and pharmacy. Follow by the steps of designing and building a computerized system. Finally, we discuss techniques for classification to the prediction of blood glucose in diabetic patients, such as linear regression, artificial neural networks and decision trees. The results obtained by them may help to define a diabetic diet, basic treatment to prevent complications. To finish, we evaluate and compare the techniques used. The regression took a rate of 57% by approximation of the true value, while the decision tree and the neural network achieved success rates of 65% and 74%, respectively, by classifying the data. Numbers that have been discussed and analyzed at the end of this paper.
Palavras-chave
Simulação (Computadores digitais), Redes neurais (Computação), Diabetes, Algoritmos de computador