One-class support vector machine based undersampling: Application to churn prediction and insurance fraud detection

摘要

In this paper, we propose One Class support vector machine (OCSVM) based undersampling. To demonstrate the effectiveness of the proposed methodology, we worked on Automobile Insurance fraud dataset and Credit card customer churn dataset taken from literature. We employed Decision Tree (DT), Support Vector Machine (SVM), Logistic Regression (LR), Probabilistic Neural Network (PNN) and Group Method of Data Handling (GMDH) for classification purpose. We observed significant improvement with respect to the Area Under Receiver Operating Characteristic Curve (AUC) over other techniques. For automobile insurance dataset, undersampling with the sigmoid kernel yielded AUC of 7605 when compared with Sundar kumar and Ravi, Vasu and Ravi, Farquad [9,12,47] with respect to Decision tree, while for Credit card customer churn dataset, undersampling with the radial basis kernel (proposed method) yielded significant performance with respect to DT (AUC 8506.5). We preferred DT over SVM (AUC 8728.5) as there is no statistically significant difference between them. Finally, we recommend DT over other classifiers as it also yields "if-then" rules, while achieving high AUC.