Hybrid Ensemble Gradient Boosting Algorithm to Predict Diabetes Health Care Analytics

Diabetes is a chronic illness characterized by high blood sugar that can cause major damage to the kidney, heart, eyes, brain system, and kidneys, among other organs. Healthcare analytics improves patient care by utilizing a variety of data analysis techniques. A machine learning model was built utilizing several machine learning approaches to predict the sickness as soon as feasible in order to prevent it. The Andaman and Nicobar Islands’ 770 diabetic individuals are the subject of the research project. In this research, the collected dataset have been splits into training and testing sets using exploratory data analysis techniques for data pre-processing. Subsequently, the study employed feature engineering methodologies to ascertain the significance of every attribute and make an accurate prediction of diabetes mellitus based on the risk factor identified by the Indian Diabetes Risk Score (IDRS). The current research work′s model employs nine machine learning algorithms, including the Gaussian Naïve Bayes algorithms, Ada Boosting classifier, XG Boosting classifier, Random Forest classifier, Bagging classifier, Logistic Regression, Linear SVC Algorithm, KNN classifier, and decision trees algorithm and produces accuracy, precision, recall, and f1 score. Depend upon the findings, it was concluded that the Bagging classifier, Ada Boosting, XG Boosting, KNN, and Random Forest classifiers provided the highest accuracy of 88%, followed by the Gaussian Naïve Bayes method, Decision Trees, and Logistic Regression. Subsequently, the research effort created a hybrid ensemble gradient boosting method and applied it to the proposed system, yielding the greatest results in terms of accuracy (99%), precision (79%), recall (84%), AUC (87%), and ROC curve (87%). This leads to the conclusion that the proposed hybrid ensemble gradient boosting classifier not only produces superior results and beats the other machine learning methods, but it also correctly predicts patients’ likelihood of developing diabetes mellitus based on risk variables.