Fraud Detection for Credit Card Transactions

Fraud credit card transaction causes severe financial damage to businesses and people. Conventional credit card fraud detection technologies rely on rigid regulation rules to figure out abnormal purchase behaviors. They are incapable of identifying complex and covert cheating crimes. This article proposes a fraud detection procedure for uncovering fraud credit card transactions. The proposed method is composed of the following steps: (1) A training data set is produced from a collection of credit card transactions. (2) The key features of transactions are identified and used for fraud detection. (3) The proposed procedure train a Self-Organizing Map (SOM) to classify normal and fraud purchases via credit cards. (4) Following the SOM training process, a decision tree is constructed to further classify normal and anomaly transactions. Along with the fraud detection process, informative visualization techniques are utilized to enhance the comprehension of features, training results, and verification of detection.