Model Accuracy Test for Early Stage of Diabetes Risk Prediction with Data Science Approach

People with diabetes should be examined periodically to create a thorough clinical record and enable preventive measures. Machine learning is the technology that is currently being preferred to perform early-stage disease detection. Although various machine learning models are available, the most accurate model for forecasting the risk of developing early-stage diabetes is still unknown. This research study compares 14 machine learning models—neural network, logistic regression, SVC, gradient boosting classifier, extra trees classifier, bagging classifier, AdaBoost classifier, gaussian NB, MLP classifier, XGB classifier, LGBM classifier, k-nearest neighbor classifier, decision tree classifier, and random forest classifier to determine the best-suited algorithm for diabetes risk prediction. From the obtained results, it is evident that the random forest with extra trees classifier has delivered the best accuracy for predicting diabetes with an accuracy rate of 99.04%. With an accuracy value of 97.12%, the gradient boosting and bagging classifier models produced the next-best result.