Asian Research Journal of Current Science

Credit card scams are susceptible to exploitation and are frequently targeted due to their inherent vulnerabilities. The proliferation of e-commerce and several other online platforms has led to an expansion of online payment methods, thereby heightening the susceptibility to online fraudulent activities. In response to the escalating incidence of fraudulent activities, scholars have turned to various machine learning techniques as a means to identify and analyze instances of fraud within the realm of online transactions. The primary objective of this study is to propose and implement an innovative approach for detecting credit card fraudulent activities from the Kaggle European cardholder dataset using a novel Deep Artificial Neural Network (DANN) by stacking multiple hidden layers on top of each other. During the process, the Linear Discriminant Analysis (LDA) algorithm and Synthetic Minority Over-Sampling Techniques Edited Nearest Neighbor (SMOTE-ENN) were used for feature selection and data balancing respectively. SMOTE-ENN is a hybrid sampling algorithm combining synthetic minority over-sampling techniques and edited nearest neighbour. To analyze the performance of the DANN algorithm various metrics, including precision, recall, f-measure, and accuracy were employed. Experimental results demonstrated that the proposed approach achieved an accuracy of 98%, a precision of 100%, a recall of 98%, and an f1-score of 99%.