Fraud detection using support vector machine ensemble

As a general fraud detection method in customer resource management user profiling can eventually be induced to binary classification and multi-classification problems of support vector machine (SVM). In this paper, we propose the SVM ensemble. in which similar SVMs are employed to detect a specific pattern of customer fraudulent behavior, and majority strategy are used to issue fraud alarm. Compared with other machine learning models, such as MLP and SOM models, either the classification accuracy or the adaptability, of the SVM ensemble is remarkable. Our application simulation shows that our design hay better performance than our previously developed system.