Imbalanced datasets often disproportionately represent certain types of malware, which can negatively impact the performance of machine learning classifiers. This imbalance can result in insufficient data for rarer but highly dangerous malware, leading to potential detection failures with serious consequences. To address this, data balancing techniques have proven effective in improving the representation of minority classes and mitigating bias toward the majority class. Recent studies have also shown that generative models can successfully create synthetic data that closely mirrors real datasets. In this paper, we explore various balancing techniques and generate synthetic opcode sequence data to enhance the training of machine learning models for improved malware classification. Our approach includes oversampling, undersampling, hybrid sampling, and the use of Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP) to generate synthetic samples. We assess the effectiveness of these methods in tackling the class imbalance problem in multi-class malware classification.