Money laundering poses a significant threat to global financial systems, enabling illicit activities such as drug trafficking and terrorism. Traditional anti-money laundering methods often fail to detect complex and evolving patterns in financial transactions. This paper explores advanced machine learning techniques to enhance the detection of suspicious activities. We evaluated the performance of six classifiers—RandomForest, LogisticRegression, KNeighbors, GradientBoosting, LightGBM, and XGBoost—on an imbalanced dataset for laundering transaction classification. Four sampling methods were employed: Without Sampling, SMOTE, RandomUnderSampler, and ADASYN. The results indicate that RandomForest, particularly when combined with SMOTE and ADASYN, consistently outperformed other classifiers. Without sampling, RandomForest achieved an F1 Score of 0.37 for the minority class (Class 1), which improved to 0.41 with SMOTE and 0.40 with ADASYN. These findings underscore the robustness of RandomForest in managing imbalanced datasets and emphasize the critical role of effective sampling techniques. This study offers valuable insights for enhancing predictive models in financial fraud detection.