Detecting anomalies in financial fraud is challenging due to data imbalance and the variety of fraudulent techniques employed. Traditional autoencoders often exhibit bias towards the majority class, reducing their effectiveness in identifying anomalies. This paper introduces an enhanced autoencoder model that improves performance on imbalanced datasets by incorporating class-specific reconstruction losses and gradient clipping techniques. To avoid potential noise and bias, we did not employ the Synthetic Minority Over-sampling Technique. Instead, we evaluated our model using the Bank Account Fraud dataset and the BankSim dataset. Our findings reveal significant improvements in key performance metrics, including precision, recall, and F1-score, compared to traditional autoencoders. These results highlight the potential of advanced deep learning techniques to enhance anomaly detection systems.