Learning-Based Fraud/Error Classification using Restatement Data

• In Event: Forensic and Investigative Accounting 1

Abstract

In this paper, we propose using machine learning techniques to distinguish errors from frauds. We treat the task as a binary classification problem, i.e., mapping from an input vector of indices to a class label of either error or fraud. We empirically evaluate and analyze five state-of-the-art classifiers, including logistic regression, artificial neural network, support vector machines, decision trees, and bagging. There are several important observations from the experimental results. First, it is observed that bagging performs the best among these commonly used general purpose machine learning tools. Second, it is very challenging to distinguish errors from frauds and general machine learning approaches, though perform better than pure chance, have large rooms for improvement.

Authors

• Mei Zhang, Rowan University

• Carol Welsh, Rowan University

• Haibin Ling, Temple University