Smart credit card fraud detection system based on dilated convolutional neural network with sampling technique

Numerous organization including financial industry are highly supported the online service payments due to the massive growth of Internet commerce and banking. But, those financial industry faces global losses due to increases of fraud and also the customer losses the trust in online banking, because credit card frauds (CCF) are mostly occurred due to high usage of Internet. Therefore, financial institutions and merchants faces the heavy losses, because illegal transactions are carried out by unauthorized user without the knowledge of actual card users. In addition, availability of public data, high false alarms, imbalance problems in data, changing nature of frauds increases the challenges in the detection of CCF. Researchers uses the Machine Learning (ML) techniques for designing the Detection system for CCF (CCFD), however, these ML didn't offers much efficiency. Therefore, to solve these issues of ML, nowadays, Deep Learning (DL) is applied in the area of CCFD. In this research work, one-dimensional Dilated Convolutional Neural Network (DCNN) is designed to solve the issues of CCFD by learning both spatial and temporal features. Here, the base model of CNN is improved by implementing the dilated convolutional layer (DCL). The imbalance problem is solved by under-sampling and over-sampling techniques. The experiments are carried out on three datasets in terms of various parameters and compared with existing CNN model. The simulation results proved that proposed DCNN model with sampling technique achieved 97.39% of accuracy on small card database, where CNN achieved 94.44% of accuracy on the same database.