A Credit Card Fraud Identification Technique Using Support Vector Machine

Transactions using credit card are increasing rapidly due to the advancements in electronic commerce. The clients as well as merchants are seriously affected by frauds in modern days and that causes high commercial losses. To minimize this loss, financial institution and banks requires a fraud detection system. Random Forests (RF) approach is method to developed for detecting fraud’s in credit card. In this method, training about abnormal and normal transactions is done using CART-based Random Forest (CARTRF) and Random-Tree-Based Random Forest (RTBRF). Complexity of this system made difficulties in the usage. This system requires large computational resources and they are harder and lea intuitive. To improve the accuracy with high true positive rate, Support Vector Machine (SVM) is proposed. From Taiwan and German dataset, default data of credit card clients are collected and preprocessed. Attributes are normalized using min-max normalization. Attribute selection based on Information Gain (IG) is used for reducing feature set. Frequent attribute selection and pruning are performed using Aprori algorithm. Candidate’s itemset size is reduced by this and improved performance gain is produced. Support Vector Machine (SVM) performs detection of frauds in credit card using frequent attribute set. The parameters like recall, precision, detection rate and pruning time are used to compare performance of proposed method against existing methods like CART-based random forest approaches and Random-tree-based random forest approach. Experimental results show better performance of proposed technique.