Imbalanced datasets or skewed datasets need sampling before the classification process to ensure that there is no class bias towards the majority class and to improve representation of minority class. In this work, four datasets that are imbalanced with different imbalanced ratios were chosen. These datasets then underwent sampling with seven different techniques (under sampling, over sampling, weighted sampling, SMOTE, Tomek links, Nearmiss and SMOTE + Tomek links: hybrid technique) and classified using four algorithms Logistic Regression (LR), Decision Tree (DT), Random Forest (RF) and KNN classification. The results from each combination of classification models with respect to its sampling techniques have been compared and analyzed using different evaluation metrics. On the Glass dataset RF with SMOTE sampling showed accuracy of 90.69%. On the PIMA dataset, DT with SMOTE sampling showed an accuracy of 77.27%. On Winequality dataset RF, LR and KNN without sampling showed an accuracy of 96.87%. On the Yeast dataset, RF with SMOTE resulted in an accuracy of 99.02%.