Optimization of Machine Learning Algorithms for Diabetes Prediction

At present, a lot of people struggle with diabetes, which is a very serious issue. Diabetes-related problems increase a person's risk of contracting diseases including coronary artery disease, kidney failure, vision difficulties harm to tissues, and other diseases. The current focus of the medical industry is gathering data on the problems related to diabetes through medical evaluations using various tests. The proper treatment is then provided based on the medical examination, which necessitates additional efforts and resources. Machine learning (ML), an emerging discipline, is essential for identifying and forecasting a wide range of events, including illness. In this study, well-known diabetes prediction methods such as Random Forests (RF), Logistic Regression (LR), XGBoost (XGB), K-Nearest Neighbours (K-NN), Decision Tree (DT, LightGBM, and Neural Network were selected. The best algorithm for the diagnostic system is found through a comparison of algorithmic capabilities. On the basis of several performance indicators, the algorithms are compared. The study's findings demonstrate that XGBoost performs better than other machine learning techniques in predicting diabetes, with the greatest accuracy of 95.5% and the lowest F1-score, recall, and precision at 92.6%, 92.7%, and 92.5%.