Comparing Different Machine Learning Approaches for Diabetic Condition Diagnosis on Balanced and Unbalanced Datasets

In order to determine whether or not machine learning models can be trusted and to pinpoint the areas where the current system falls short, several approaches are evaluated. This research proposes using or developing new ways of modeling for diabetes diagnosis as the best method accuracy is low. The availability of a suitable dataset is the main obstacle for computer-assisted diagnostics. There are insufficient datasets with a balanced label for studying diabetes. This work attempts to demonstrate the value of balanced datasets by putting them into practice and contrasting the results with those obtained by modeling with an uneven distribution of data. Random Forest produces the best outcomes for both balanced and unbalanced data.