MÉTODOS DE CLASSIFICAÇÃO AUTOMÁTICA PARA PREDIÇÃO DO PERFIL CLÍNICO DE PACIENTES PORTADORES DO DIABETES MELLITUS

Abstract

The goal of this paper is to study the relationships between the main attributes that influence the diagnosis and control of Diabetes Mellitus Type 2 and to generate an automatic classification tool that allows inferring about the glycemic index and which can be used as a medical aid in order to the patient with diabetes can be directed to the appropriate treatment. The methods proposed for this task are based on Bayesian Classification method, which uses the BayesRule algorithm and is able to investigate probabilistic uncertainties in the data, and on the classification method using Decision Trees, which is a classification tool widely used in data mining due to easy interpretation of the results. Both methodologies extract linguistic classification rules, which allows the comparison of their performances. According to the cross-validation process, the Bayesian classification method with the BayesRule algorithm presents 65% accuracy in the classification task for the intervention group and 47.5% for the control group. The Pruning Decision Trees present 73.68% accuracy for the intervention group and 69.23% for the control group. Then the results obtained in this study are satisfactory, and may contribute to the control and prediction of the development of patients with Diabetes Mellitus Type 2.