In foresight: Detecting financial statement fraud probabilities and extracting indicators through artificial intelligence methods

• المؤلفون الرئيسيون: Batungbacal, Esther Grace T., Murakami, Criselle A., Shi, Ya Ling, Sy, Mary Ann S.
• التنسيق: text
• اللغة: English
• منشور في: Animo Repository 2015
• الموضوعات: Misleading financial statements; Financial statements; Accounting fraud
• الوصول للمادة أونلاين: https://animorepository.dlsu.edu.ph/etd_bachelors/8410
• المؤسسة: De La Salle University

Financial statement fraud is the deliberate misstatement and misrepresentation of data to mislead the readers and conceal the financial strength of an organization. In the past decade, numerous accounting scandals transpired with devastating consequences. Enron Corporation (2001), WorldCom (2002), and Tyco (2002) are just three of the most prominent cases of financial statement fraud. Studies conducted by the Association of Certified Fraud Examiners show that over the years, financial statement fraud has evolved, not just in Europe and America, but in Asia as well. It is a pressing concern as huge amounts of losses are incurred in the countries on a yearly basis. With this, it is apparent that immediate action must be taken. As fraudulent practices prevail in Asia, this paper looks into the financial statements of seventy-six (76) companies belonging to the Philippines, China, Singapore, Japan, and Malaysia and employs four (4) artificial intelligence methods: Bayesian belief networks, neural networks, decision tree, and logistic regression. The purpose of this study is to identify the most accurate detection model, and from the model, indicators that will serve as red flags are to be extracted. From the method used, neural networks appeared to be the most accurate model. From the most accurate model, the common indicators extracted that exhibit a strong relationship with fraud were inventory over total assets (INVTA) and logarithm of total debt (LOGDEBT). The results of this study show that artificial intelligence methods can be used in a broader spectrum as the methods are flexible with other countries and various industries. This research is of positive contribution in providing a fraud detection model that can be utilized by professionals, investors, government, financial institutions, or any other related parties to mitigate potential losses and lessen risks of investing in fraudulent entities.