Improving Diabetes Prediction Accuracy Using Wb-Smote for Imbalanced Datasets

Serious morbidity and early death are consequences of diabetes, a worldwide disease. Several organs are damaged by it, including the kidneys, heart arteries, eyes, and eyeballs, which leads to a rise in yearly death rates. To tackle this tendency head-on, we urgently need an advanced automated prediction system that can identify illnesses early. Producing a medical administrative model that can reliably foretell diseases via the use of ML methods is the primary goal of this research. Improving decision-making skills, transparency, and accuracy are just a few of the benefits of using computer-based algorithms. An very accurate model for predicting the likelihood of acquiring diabetes is the goal of the research. For better preprocessing methods and to fix imbalanced data, the WB-SMOTE method was created. To address datasets with class imbalances, the WB-SMOTE approach was developed independently. We include the Borderline SMOTE methodology, which chooses the oversampling of each minority dataset according to pre-set weights, to improve the algorithm. Our method handles common minority concerns and gives border points more weight than noise points while creating synthetic data. The impoverished minority class is strengthened by the resulting paradigm. We must first normalize the minority class. After that, the input dataset is supplemented with synthetic data that is created using boundary points.