<![CDATA[Diabetes is a chronic disease with a globally increasing prevalence and poses a serious health risk if not detected early. This study aims to develop a simple yet accurate classification model to predict diabetes risk using the K-Nearest Neighbors (KNN) algorithm. The model utilizes four key clinical parameters: body mass index (BMI), blood glucose level, glycated hemoglobin (HbA1c), and age. The dataset used in this study was obtained from Kaggle and consists of 50,066 records. The data underwent preprocessing stages including normalization, class balancing, and feature selection before being split into training (70%), testing (20%), and validation (10%) sets. Experimental results demonstrate that the KNN model with k=4 achieves an accuracy of 83% on the validation set and 82% on the test set. Although the model shows stable and reasonably good performance, further improvement is needed to achieve better class balance in prediction. Future work may involve exploring more advanced machine learning techniques to enhance predictive capabilities and ensure fair classification across different risk groups.]]>
Copyrights © 2026
