Selection Features and Support Vector Machine for Credit Card Risk Identification

Abstract For identifying credit card risk in massive and high dimensionality data, feature selection is considered very important to improve classification performance and fraud identification process. One of the commonly used feature selection methods is Random Forest Classifier (RFC), which is very suitable for large dataset. RFC has a good performance; it tends to identify the most predictive features, which may provide a significant improvement in classification performance of credit card risk identification model. In this paper, we propose an enhanced Credit Card Risk Identification (CCRI) method based on the features selection algorithm as Random Forest Classifier and Support Vector Machine to detecting fraud risk. Our experimental results show that the proposed algorithm outperforms the Local Outlier Factor, Isolation Forest and Decision Tree in term of classification performance on a larger dataset.