Detecting insurance fraud is one of the most important problems for the insurance industry. While several methods have been proposed, no method provides provably valid predictions. Based on a general machine learning strategy, we propose a new method for detecting insurance fraud that guarantees finite-sample validity. The proposed method is distribution-free, tuning-parameter-free, and applicable to both continuous and categorical features. It can also be used to detect other types of fraud such as credit card fraud.
Yang et al. (Wed,) studied this question.