An Early Diabetes Diseases Prediction Using Machine Learning With Optimal Features Selection
Prabakaran Kasinathan, M. Sangeetha, M. Balamurugan, Shaik Aminabee, Beema Rao, Pundru Chandra Shaker Reddy
Abstract
Diabetics have elevated blood glucose levels for a lengthy period of time, either due to insufficient insulin synthesis or a lack of effective insulin response by the body's cells. Long-term harm, breakdown, and collapse of diverse organs, including the eyes, kidneys, nerves, heart, and veins, are associated with the persistent hyperglycemia of diabetes. The focus of this study is on utilizing important features, developing a prediction algorithm by means of Machine-learning(ML), and identifying the best classifier in order to obtain the most accurate results when compared to clinical outcomes. The proposed approach utilizes Predictive analysis with the goal of zeroing in on the variables that are crucial for early diagnosis of Diabetes Miletus. There were a variety of methods used, including random-forest(RF), extreme-gradient-boost(EDB), logistic-regression(LR), and weighted-ensemble-models(WSM). When compared to severe gradient boost (0.93), logistic regression (0.92), and weighted ensemble model (0.87), the RF performed admirably in anticipating uncontrolled diabetes. The study also generalizes the process of picking the best features from the dataset to boost classification precision. These clinical features can be utilized in conjunction with machine learning methods for diabetes control prediction.