The Comparison of Two Data Mining Method to Detect Financial Fraud in Indonesia

Abstract

Objective - This research is expected to improve the weaknesses in the research conducted by Tarjo and Herawati (2015). The objective of this study was to analyse two data mining methods in detecting financial fraud based on Beneish m-score model.

Methodology/Technique - The research data were companies who committed fraud based Database Case Sanctions Issuers and Public Companies which was released by the Financial Services Authority in the period 2001-2014. For comparison, researchers also used data from companies that did not commit fraud. Companies were selected based on the same industry group of companies committing fraud for the purposes of classification.

Findings - The results show that data mining methods can be used to detect financial fraud based on Beneish m-score model. However, there are differences in the classification. In the logit regression, the results are only limited to the accuracy of classification and weak. While the K-Nearest Neighbor model, in addition, it is capable of performing high classification accuracy.

Novelty - The study indicates a better method for detecting financial fraud.

Keywords: Detecting Financial Fraud, Beneish M-Score, Logit Regression, K-Nearest Neighbor

JEL Classification: C81, M41