Prediabetes Detection Using Non-Laboratory Data with Extreme Gradient Boosting Algorithm

Diabetes Mellitus (DM), characterized by high blood glucose levels, is a dangerous condition that leads to severe long-term health issues. Meanwhile, prediabetes is a condition where individuals have an increased risk of developing type 2 DM. Early detection of prediabetes is crucial for effective DM management. Traditional detection relies on medical records and family history, but laboratory tests are not always practical. This study proposes using non-laboratory data, processed by the Extreme Gradient Boosting (XGBoost) algorithm, to identify individuals with prediabetes. The dataset used in this research includes attributes like blood pressure, weight, height, waist size, food intake, and age. Preprocessing involved removing irrelevant data and transforming numeric data into categories. The XGBoost model, applied to this dataset, achieved 0.91 for average accuracy, precision, recall, and F1-score. AUC values of the model is 0.93, demonstrating its effectiveness in distinguishing between normal and prediabetic individuals. These results suggest the potential for non-laboratory data and XGBoost in early prediabetes detection, aiding in timely intervention and improved health outcomes.