Abstract:
Banks provide their services to make money transactions and other money-based banking activities. The basic concept of the bank is the transaction of money; it can be either manual or digital through multiple channels. Currently, electronic-based money transactions are more popular among customers. Due to the advancement in technologies, perpetrators also get a chance to steal money in unauthorized ways. This study considers fraud transactions among credit card transaction history as a digital forensic footprint to analyze and detect fraud transactions using machine learning techniques. Therefore, a dataset from an American Bank, USA, was used to build a model for fraud detection. The dataset consists of genuine and fraud transactions having 31 attributes each. The dataset was preprocessed, and the statistical features were extracted by applying different sampling techniques. Further, some combination of feature sets is used with different classifiers to create a better model. Classification methods were analyzed through confusion matrix and execution time. Finally, it originates from a suitable model to recognize fraud detection with credit card transaction data. The predicted model produces 99.999% of accuracy in detecting fraudulent transactions using a random forest classifier.
