International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Diabetes, commonly known as diabetes mellitus, is a condition that affects how the body processes blood sugar. It occurs when the pancreas either cannot produce enough insulin or the body is unable to effectively use the insulin that is produced. Insulin, a hormone secreted by the pancreas, facilitates the transport of glucose from food into cells, where it is used for energy. Uncontrolled diabetes often leads to hyperglycemia (high blood sugar), which, along with other health complications, can significantly damage nerves and blood vessels. According to 2014 statistics, a substantial number of individuals aged 18 and older had diabetes, and in 2019, diabetes alone was responsible for 1.5 million deaths.

However, with the rapid advancement of machine learning (ML) and deep learning (DL) classification algorithms, early detection of diabetes has become significantly more feasible across various fields, including healthcare. In this study, we conducted a comparative analysis of multiple ML and DL techniques for early diabetes prediction. We utilized a diabetes dataset from the UCI repository, comprising 17 attributes, including the target class, and evaluated the performance of all proposed algorithms using a range of performance metrics. Our experiments indicated that the XGBoost classifier outperformed all other algorithms, achieving nearly 100% accuracy, while the remaining models demonstrated accuracy levels exceeding 90%.

Keywords: Diabetes prediction; XGBoost; KNN; CNN; LSTM; Classification.

Downloads: | DOI: 10.17148/IJARCCE.2025.14820