It is imperative to address credit card fraud to protect consumer trust and enhance financial security. The primary goal of this paper is to design a reliable system which is capable of detecting credit card fraud or non-fraud with the utmost accuracy. The deep multilayer perceptrons, is used to facilitate the precise prediction of complex patterns and the accurate identification of the classes of a dataset. The proposed model consists of 12 consecutive layers, to acquire intricate patterns from transactional data with a high number of dimensions. The utilization of key approaches like feature scaling, class weighting, and dropout regularization is implemented to improve the performance and reduce the occurrence of overfitting. The performance of the model measures using a dataset of credit card transactions by using different split ratios and a 10-fold cross-validation process, with considerable imbalance between fraudulent and non-fraudulent activities. The goal of this model is to help the improvement in the detection of fraudulent activities, as evidenced by measures such as accuracy, precision, recall, and f1-score.