Credit Card Fraud Detection Using Symbolic Aggregate Approximation

In the current digital age, credit card fraud is a cybercrime that costs billions of dollars every year through fraudulent transactions. This study aims to identify fraudulent credit card transactions using Symbolic Aggregate Approximation (SAX) and anomaly discovery through motifs and discords. To do this, credit card data is used as time series data and then divided into equal-length segments. Each segment is then converted into a symbolic representation using a pre-defined threshold. A web-based interface is further developed to load a collection of credit card transaction data, based on the proposed SAX method. The preliminary results show that the proposed approach is capable of identifying fraudulent transactions within a set of transactions. By integrating this approach with other fraud detection mechanisms, a near real-time fraud detection process can be developed. This approach therefore provides a potential for context-independent fraud detection which is a major challenge in the fraud detection domain.