سیگنال های هشدار دهنده برای تقلب های حسابداری بالقوه در شرکت های تراشه آبی - استفاده از تئوری رزونانس تطبیقی

With the constantly changing and deceptive strategies that can be concealed in complex of financial statements, traditional means of financial analysis is unable to detect these accounting frauds in advance. In order to detect new accounting frauds and find out the true meaning of off-balance sheet arrangements, we propose an easy and feasible method using an unsupervised learning system. In unsupervised learning, the training of the network is entirely data-driven and no target results are provided. The features that do not help in clustering can be removed. With unsupervised learning it is possible to learn larger and more complex relations than with supervised learning. In the demonstration, we extract four non-traditional warning signals using adaptive resonance theory, with Enron and WorldCom as prototypes to identify the possibility of potential fraud of a company that investors or analysts may be concerned with.

Although there are certain companies investors consider to be very safe, the confidence has been shaken by a number of accounting frauds in many large American companies, ranging from hospital suppliers to underwater optical cable and security systems. Large investment banks and independent public accountants have also been involved. In order to secure business from the companies concerned, investment banks present false advice to the public. When the consulting services become very lucrative, there is increasing conflict of interest between the auditing role and consulting role. Because financial statements are the basis for measuring a company’s performance, accounting frauds not only damage the confidence of investors but also damage all the management analysis related to performance. The Sarbanes-Oxley Act of 2002 was imposed to deal with some of these serious frauds, but the suggested reforms still fail to address the major problem that chief executive officers (CEOs), chief financial officers (CFOs), and other directors and officers serve their personal interests to the detriment of shareholders. With the esoteric and deceptive strategies in the complexity of financial statements, the traditional forms of financial analysis are unable to detect versatile accounting frauds. In order to detect these accounting frauds and to find out the true meaning of off-balance sheet arrangements, we propose an easy and feasible method using an unsupervised learning system. We use the factor of whether or not significant motive or pressure exists as the complement coding. Even if there are a lot of complex off-balance sheet arrangements, for the frauds involving millions or billions of losses, some items must have abnormal changes or there must be serious liquidity problems. Based on these conjectures and an unsupervised learning system, we sketch out four warning signals. With unsupervised learning it is possible to learn larger and more complex relations than with supervised learning. The training is entirely data-driven and no target results are provided. The features that do not help in clustering can be removed. The qualitative match-based unsupervised learning of adaptive resonance theory (ART) allows memories to change only when input from the external world is close enough to internal expectations, or when something completely new occurs. This feature makes the ART system well suited to problems that require on-line learning of large and evolving databases, so that it can be used for the analysis of accounting frauds.

Due to the complex and deceptive strategies that can be embedded in financial statements, traditional financial analysis may not provide a true picture of a company’s performance. In order to detect versatile accounting fraud and to find out the actual meaning of the off-balance sheet arrangements, we propose using ART and databases to extract non-traditional warning signals. In this demonstration, the warning signals extracted are: (1) the change rate of an item exceeds 50% without acceptable reasons; (2) the ratio of cash and cash equivalents divided by current liabilities is less than or equal to 25%; (3) the ratio of net tangible assets divided by total liabilities is less than or equal to 25%; and (4) the ratio of interest expense divided by cash and cash equivalents is greater than or equal to 50%. The results are the same for different values of vigilance. After learning and clustering, Enron and WorldCom are in one category. Intel, Texas Instruments, Advanced Micro Devices, STMicroelectronics, and Analog Devices are in the other category. With Enron and WorldCom as prototypes, there is no possibility of potential fraud for the other five companies. Because accounting frauds are versatile, we can use the ART system to learn on-line from databases in an unsupervised mode. The databases contain restatement files, administrative proceeding files, suspension files, and litigation files. The match-based ART can also modify warning signals by using new data and change vigilance based on prediction errors