This paper describes the ways in which forensic credit card fraud detection can be achieved using deep neural network (DNN). The problem that this research addresses is to identify a technique that can be used by forensic experts in credit card fraud investigations to detect if a fraudulent transaction has occurred. It is an obligation that institutions providing financial services must implement relevant safeguards against credit card fraud to prevent possible loss of their investments or clients’ funds. This paper presents a forensic detection model of credit card fraud that is based on sequential data modeling using Long Short-Term Memory (LSTM) DNNs. The current study determines whether LSTM-attention algorithm can identify the most important transactions in an input sequence that provides high-accuracy prediction of fraudulent transactions. The effectiveness of the LSTM-attention model is achieved by selecting the most relevant predictive features, uniform manifold approximation, transaction sequences, and attention mechanisms that improve the performance of the model. The results show that LSTM-attention algorithms can be used to conduct forensic credit card fraud detection with high accuracy and precision. The novelty of the research paper is that it successfully uses an LSTM-attention algorithm to detect credit card fraud instances and proves the applicability of the model in mitigating fraudulent transactions in banking institutions. Keywords: Credit Card Fraud, Machine Learning, Deep Neural Network, Convolutional Neural Network, Long Short-Term Memory DOI: https://doi.org/10.35741/issn.0258-2724.58.1.33