Improving the Methodology for Integrated Testing of Journal Entries by Benford’s Law

This paper explores the growing prevalence of Benford’s Law as a statistical method to identify intentional manipulations of numerical data. The study focuses on improving a methodology for applying Benford’s Law tests in detecting distortions within accounting practices. Primary, advanced, and associated tests are conducted to assess the natural character of journal entries of a construction company. Additionally, machine learning techniques such as K-means clustering, random forest, and elliptic envelope are used to analyze the test results and identify highly suspicious transactions within the dataset. The outcomes indicate that the selected transactions flagged by the tests are indeed suspicious, with significant monetary values.