Measuring the Impact of Imputation in Financial Fraud

摘要

In recent years, data mining techniques have been used to identify different types of financial frauds. In some cases, the fraud domain of interest contains data with missing values. In financial statement fraud detection, instances which contain missing values are usually discarded from the analysis. This may lead to crucial information loss. Imputation is a technique to estimate missing values and is an alternative to case-wise deletion. In this paper, a study on the effectiveness of imputation is taken using financial statement fraud data. Also, the measure of similarity to the ground truth is examined using five distance metrics.