Bank account fraud is a major issue that impacts millions of people and causes huge losses for financial institutions. The research explores the use of predictive classifiers, namely Random Forest and Gaussian Naive Bayes, to detect fraudulent account opening requests. A synthetic dataset is employed to simulate various features and scenarios of account fraud, including factors such as name-email similarity, income group, credit risk score, and payment type. Data preprocessing, cleaning, and visualization are conducted to prepare the data for modeling. Subsequently, on the test data, the performance of the two classifiers is compared using measures such as accuracy, precision, recall, and F1-score. The result of performance metrics Area under the curve (AUC) of Receiver Operating Characteristics (ROC) Curve, True positive rate (TPR), False positive rate (FPR), and threshold for Random Forest are following in order 0.8985, 0.5759, 0.4310, 0.04 and for Gaussian Naïve Bayes 0.8789, 0.5279, 0.0498, 0.74. The feature importance and confusion matrix of the models are also analyzed. Notably, Random Forest demonstrates superior performance to Gaussian Naive Bayes across most metrics. Additionally, certain features, such as name-email similarity and velocity, emerge as more indicative of fraud than others.