Support Vector Machine Approach for Detecting Credit Card Frauds

Financial fraud is increasing significantly with the development of modern technology and the global superhighways of communication, resulting in the loss of billions of dollars worldwide each year. The companies and financial institution loose huge amounts due to fraud and fraudsters continuously try to find new rules and tactics to commit illegal actions. Thus,  fraud detection systems have become essential for all credit card issuing banks to minimize their losses.  The most commonly used fraud detection methods are Neural Network (NN), rule-induction techniques, fuzzy system, decision trees, Support Vector Machines 

(SVM), Artificial Immune System (AIS), genetic algorithms, K-Nearest Neighbor algorithms.

The detection of fraud is a complex computational task and still there is no system that surely predicts any transaction as fraudulent. They just predict the likelihood  of the transaction 

to be a fraudulent.

The properties of a good fraud detection system are:

1) It should identify the frauds accurately

2) It should detecting the frauds quickly

3) It should not classify a genuine transaction as fraud

Support Vector Machine Approach

The Support Vector Machines (SVM) is statistical learning techniques and has successful application in a range of problems. It was first introduced by Cortes and Vapnik (1995) and it has been found to be very successful in a variety of classification tasks. They are closely related to neural networks and through the use of kernel functions, they can be considered an alternative way to obtain neural network classifiers. SVM algorithm is a supervised machine learning algorithm that has been applied to anomaly detection in the one-class setting. Such techniques use one class learning techniques for SVM and learn a region that contains the training data instances. The basic idea of SVM classification algorithm is to construct a hyper plane as the decision plane which making the distance between the positive and negative mode maximum. The strength of SVMs comes from two important properties they possess  -kernel representation and margin optimization. Kernels, such as  radial basis function (RBF) kernel, can be used to learn complex regions. A kernel function represents the dot product of projections of two data points in a high dimensional feature space. In SVMs, the classification function is a hyper-plane separating the different classes of data. The basic technique finds the smallest hypersphere  in the kernel space that contains all training instances, and then determines on which side of hypersphere a test instance lies. If a test instance lies outside the hypersphere, it is confirmed to be suspicion. This algorithm finds a special kind of linear model, the  maximum margin hyper plane, and it classifies all training instances correctly by separating them into correct classes through a hyper plane. The maximum margin hyper plane is the one that gives the greatest separation between the classes. The instances that are nearest to the maximum margin hyper plane are called support vectors. There is always at least one support vector for each class, and often there are more. In credit card fraud detection, for each test instance, it determines if the test instance falls within the learned region. Then if a test instance falls within the learned region, it is declared as normal, else it is declared as anomalous. This model has been demonstrated that it possess a higher accuracy of detection compared with other algorithms. It also has a better time efficiency and generalization ability. Performance evaluation of SVM with BPN in credit card fraud detection shows that when the data number is small, SVM can have better prediction performance than BPN in predicting the future data. But in large data BPN has a good performance.

References:

[1] Lean Yu a, Wuyi Yue, Shouyang, Wang, K.K. Lai “Support vector machine based multiagent ensemble learning for credit risk  evaluation”. Expert Systems with Applications(2010).  37; (1351–1360).

[2] N. Cristianini, J. Shawe-Taylor “An Introduction to Support Vector Machines and Other Kernel-based 

Learning Methods”. Cambridge University Press. (2000).

[3] Qibei Lu, Chunhua Ju “Research on  Credit Card Fraud Detection Model Based on Class Weighted Support Vector Machine”.  Journal of Convergence Information Technology, (2011).  Volume 6, Number 1; (62-68).