Financial fraud detection: A new ensemble learning approach for imbalanced data

With the rapid development of online and offline transactions, various financial fraud crimes happen every day. Financial fraud has seriously affected the health of economics and damaged the welfare of consumers, investors, as well as financial institutions. Prior studies apply several classification technologies, including decision trees, Bayesian networks, and support vector machines (SVM), to detect fraud detection. However, they ignore one important characteristic of fraud data, which is the number of valid records is largely smaller than the number of illegal fraud records. It implies the data is imbalanced. To resolve this issue, some researchers combine different sampling techniques to improve the detection accuracy of imbalanced fraud data. Among these techniques, ensemble learning is regarded as a perfect tool to handle the classification in imbalance data set. In this study, we propose a new ensemble method for financial fraud detection. This approach combines the bagging and boosting techniques together, in which the bagging technique can reduce the variance for the classification model through resampling the original data set, while boosting technique can reduce the bias of the model. In the future, we would conduct a series of experiments to evaluate the effectiveness of our approaches with the other state-of-the-art methods on real datasets.