Comparison of Poisson process and machine learning algorithms approach for credit card fraud detection

摘要

This article describes the financial fraud detection in imbalanced data. We compare various approaches for credit card fraud detection problem. On the one hand, we use homogeneous and heterogeneous Poisson process to determine the probability of predicting fraud with the various intensity parametric functions. On the other hand, we solve classification problem using machine learning algorithms and different family of ensemble methods like boostings. The results of both methods are compared. The “false positive” problem is also discussed in the article.