In healthcare machine learning is used mainly for disease diagnosis or acute condition detection based on patient data analysis. In the proposed work diabetic patient dataset analysis is done for hypoglycemia detection which means the lowering of blood glucose level (BGL). Often in healthcare it is observed that the dataset is imbalanced. Therefore, an Ensemble Approach using imbalanced dataset techniques synthetic minority over-sampling technique and adaptive synthetic oversampling methods with different evaluation methods like train-test, K-fold, stratified K-Fold and repeat traintest were used. This ensemble approach was implemented on diabetic dataset using K-nearest neighbor (KNN), support vector machine (SVM), random forest (RF), Naïve Bayes (NB) and logistic regression classifiers with average Stacking-C method thereafter to conclude. Comparative analysis was done using three different considerations. The results showed that KNN and random forest gives more stable metric values both on balanced and imbalanced dataset. The confusion matrix consideration concluded that KNN and random forest were found to be better with least false negative and maximum true positive count. But if average train and test time is taken into consideration then Naïve Bayes and random forest had least average traintest time. Thus, the three different considerations concluded that the proposed ensemble approach gives better clarity for different classifier implementation using machine learning.