Predicting Type 2 Diabetes based on Machine and deep learning models

Abstract

The global rise of spread the Type 2 Diabetes (T2D) disease has become a major global health challenge. Early detection of this disease is essential for limiting its progression and reducing its potential effect. This study estimates the performance of machine learning (ML) and deep learning (DL) models for prediction T2D using an enhanced version of Pima Indian diabetes dataset. The dataset has been improved through comprehensive preprocessing, feature engineering, and class imbalance handling via Synthetic Minority Oversampling Technique (SMOTE). A total of Seven machine learning classifiers- Logistic Regression, Support Vector Machine, Random Forest, Decision Tree, k-Nearest Neighbors, Naïve Bayes, and XGBoost were assessed alongside three deep learning models Artificial Neural Network (ANN), Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). The Experimental results demonstrate that, XGBoost achieved best predictive performance, with an accuracy (96.84\%, AUC = 0.99), followed by Random Forest (96.32%, AUC = 0.98). Decision Tree and SVM also showed robust performance, while Naïve Bayes was the least accurate 80.5%. In contrast, The DL models achieved an accuracy between (73–77%).