Authors :

Fatima Al Mansoori and Laila Mohammed Al Qubaisi

Address :

Faculty of Department of Computer Science, Khalifa University, UAE

Department of Biomedical Sciences, UAE University, UAE

Abstract :

Millions of people throughout the world suffer from diabetes mellitus, an inflammatory disorder characterized by consistently elevated blood glucose levels. Among the numerous consequences that can arise from diabetes include nerve and kidney problems, heart disease, and retinopathy. Medical professionals can slow the disease's course and lessen its impact when caught and treated early. The study recommends a Predictive Data-Mining Platform for Presymptomatic Diabetic Risk Exploration (PreDiX) to determine whether presymptomatic people may be reliably predicted to develop diabetes using data-mining techniques. To help with early risk assessment, the study used a large dataset that included demographics, way of life, and health-related factors to find patterns and connections. Examining the predictive power of data mining algorithms, including logistic regression, decision trees, Naive Bayes, and random forest modelling for diabetes occurrence, is the explicit goal of the study. The study assessed these models using the conventional criteria of accuracy, precision, recall, and area under the receiver operational features curve (AUC-ROC). Improving public health and decreasing the strain of diabetes on medical facilities are potential results of this study. Possible results of the research include intervention programmes that are both specific and preventive.

Keywords :

Presymptomatic diabetes, Predictive modelling, Early detection, Random Forest, Data mining, Risk assessment, Decision trees.