Classification is one of the critical task in datamining. Many classifiers exist for classification task and each have their own pros and cons. It is observed that due to imbalancing in datasets quality of classification accuracy is decreasing. Thus the increasing rate of data diversity and size decreases the performance and efficiency of classifiers. Thus it is very much important to get the maximum classification accuracy. Ensemble learning is a simple, useful and effective meta-classification methodology that combines the predictions from various classifiers. In this research an empirical study hasbeen done using voting based ensemble learning technique on varying imbalance data and varying organization for improving classification accuracy.