Key points are not available for this paper at this time.
This study aims to develop a comprehensive solution for predicting diabetes using machine learning (ML) algorithms. Due to rising cases of diabetes worldwide, early detection is imperative for effective health management and prevention of their complications. The objective of this research is to construct a strong model capable of accurately predicting diabetes based on diverse health data that are included in the Pima Indians Diabetes Dataset (PIDD). We explore multiple ML models, such as SVM, KNN, Bagging-DT, RF, and XGBOOST to enhance diabetic predictive accuracy. In addition, we leveraged the concept of Feature Set Creation to optimize ML model performance. Feature Sets are created by considering the various combinations of diabetic attributes. Based on this Feature Set creation, we emphasize the importance of accuracy, efficiency, and adaptability while selecting ML models for healthcare applications. Our proposed classifier, utilizing the XGBoost algorithm, demonstrates outstanding performance by achieving an accuracy of 95.29% sensitivity of 89.28%, specificity of 98.2%, and a notable F1 score of 92.59% compared with other classifiers. Additionally, a user-friendly web page has been developed to allow them to enter their BMI, age, insulin levels, and other relevant data to determine their likelihood of being diabetic or not. This website also provides specific diet recommendations for diabetic patients, emphasizing the significance of customized food plans in controlling their blood sugar, enhancing health conditions, and minimizing associated diabetic complications.
Rao et al. (Fri,) studied this question.