One of the most common cybercrimes that people encounter is credit card fraud. Systems for identifying fraudulent transactions that are based on intelligent machine learning are particularly successful in real-world situations. Nevertheless, when creating these systems, machine learning algorithms face the issue of imbalanced data or an unbalanced distribution of classes. Because of this, balancing the dataset becomes a crucial sub-task. A review of cutting-edge methods highlights the necessity for a thorough assessment of class imbalance management techniques in order to create a smart and effective system to identify fraudulent transactions. The goal of the current study is to compare several strategies for dealing with class imbalance. Therefore, the present study compares the performance of our novel K-CGAN method with SMOTE, B-SMOTE, and ADASYN in terms of Recall, F1-score, Accuracy, and Precision. The result shows that novel K-CGANs generated high quality test dataset and performs better as compared to other resampling techniques.