مدل‌سازی فرار مالیاتی معاملات اشخاص وابسته رویکرد هیبریدی گراف کاوی و شبکه عصبی عمیق

فرار مالیاتی مبتنی بر معاملات وابسته یک استراتژی جدید در فرار مالیاتی است که از طریق معاملات قانونی،مانندمعاملات بین گروهی از‌شرکت‌ها که روابط تعاملی ناهمگن،پیچیده وپنهانی برای فرار مالیاتی دارند، انجام می‌شود. مطالعات موجود نمی‌توانند به طور موثر رفتارهای فرار مالیاتی اشخاص وابسته را شناسایی کنند، زیرا روش حسابرسی مبتنی بر یادگیری ماشین می‌تواند وضعیت مالی غیرعادی افراد را با دقت و کارایی بالا تشخیص دهد. با این حال، هنگام مواجهه با روابط تعاملی ناهمگن، پیچیده و پنهانی درمانده می‌شود و نمی‌تواند گروه‌های فرار مالیاتی دارای معاملات اشخاص وابسته را شناسایی کند. هیبرید رویکردهای گراف‌کاوی و شبکه عصبی عمیق، توانایی تشخیص ناهنجاری در ساختارهای سازمانی پیچیده را دارد. در این پژوهش تعداد1780شرکت دارای معاملات وابسته، شامل 523 شرکت واقع در مناطق آزاد تجاری و 1257 شرکت واقع در خارج از مناطق آزاد که دارای عضو هیات مدیره مشترک و فعالیت اقتصادی تولیدی یا بازرگانی بوده اند، انتخاب شده‌اند. دراین پژوهش، داده‌های مالی و مالیاتی سال‌های 1395 تا 1399 ازاظهارنامه‌های مالیاتی و سامانه‌های سازمان امور مالیاتی کشور مورد استفاده قرار گرفته است. این پژوهش از نظر هدف، کاربردی می‌باشد.جهت برآورد مدل از نرم افزار پایتون و پکیج networkx بهره گرفته شده است. جهت پیش‌بینی فرار مالیاتی معاملات اشخاص وابسته از سه الگوریتم شبکه عصبی پیچشی (cnn)، حافظه کوتاه‌مدت ماندگار(lstm)،و شبکه عصبی پرسپترون چند لایه(mlp) در حالت عمیق بهره گرفته شد. برای شناسایی گروه‌های مشکوک سه مرحله؛ اول: تشخیص تفاوت نرخ مالیات، تطبیق الگوی توپولوژیکی و شناسایی ناهنجاری بار مالیاتی؛ دوم: آزمایش‌های تجربی بر اساس داده‌های 16،756 مبادله خرید و فروش معاملات وابسته درکشور؛سوم:برآورد ضرایب و نحوه ارتباط مابین الگوی توپولوژیک دردو حالت حفظ سودوانتقال سودبراساس رویکرد گراف کاوی وشبکه عصبی عمیق صورت پذیرفته است. نتایج به دست آمده نشان می‌دهد که هر دو حالت حفظ سود و انتقال سود در فرار مالیاتی معاملات اشخاص وابسته وجود داشته است. با این وجود بر اساس نتایج، شدت رابطه حفظ سود در فرار مالیاتی معاملات اشخاص وابسته قوی‌تر از رابطه انتقال سود است و نیز رویکرد گراف کاوی نسبت به مدل‌های لاجیت، پرابیت و احتمال خطی از دقت بالاتری برخوردار می باشد. 

tax evasion modeling of related party transactions ahybrid approach of graph mining and deep neural network

tax evasion based on related party transactions is a new strategy in tax evasion that is carried out through legal transactions, such as transactions between a group of companies that have heterogeneous, complex, and hidden interaction relationships for tax evasion. existing studies cannot effectively identify tax evasion behaviors of related parties because the machine learning-based audit method can detect the abnormal financial status of individuals with high accuracy and efficiency. however, it is helpless when faced with heterogeneous, complex, and hidden interaction relationships and cannot identify tax evasion groups with related party transactions. the hybrid of graph mining and deep neural network approaches has the ability to detect anomalies in complex organizational structures. in this study, 1,780 companies with related party transactions, including 523 companies located in free trade zones and 1,257 companies located outside free trade zones, which have a common board member and economic activity of production or trade, were selected. in this study, financial and tax data from tax returns and the systems of the iranian tax administration from 2016 to 2019 were used. this study is practical in terms of purpose. python software and the networkx package were used to estimate the model. to predict tax evasion in related party transactions, three algorithms were used: convolutional neural network (cnn), long short-term memory (lstm), and multilayer perceptron neural network (mlp) in deep mode. to identify suspicious groups, three steps were taken; first: detecting tax rate differences, matching the topological pattern, and identifying tax burden anomalies; second: experimental tests based on data from 16,756 related party transaction purchases and sales in the country; third: estimating the coefficients and the relationship between the topological pattern in the two cases of profit retention and profit transfer based on the graph mining approach and deep neural network. the results show that both profit retention and profit shifting exist in tax evasion of related party transactions. however, based on the results, the intensity of the profit retention relationship in tax evasion of related party transactions is stronger than the profit shifting relationship. based on the results, the graph mining approach was more accurate than the logit, probit, and linear probability models.introductiontax evasion is a financial crime against the tax system of countries in which taxpayers intentionally report false financial status to evade their tax obligations; therefore, this issue has become a serious economic problem for many countries due to the creation of a significant tax gap in the implementation of tax policies and the subsequent less reliance of governments on tax revenues in budgeting. in iran, according to some existing laws, perpetrators of tax crimes are sentenced to the penalties prescribed by law. it should be noted that recent research has used data analysis techniques to analyze and detect the tax evasion behaviors of individual taxpayers. however, they have failed to support the analysis and exploration of new tax evasion strategies through related party transactions (e. g. , transfer pricing) in which a group of taxpayers is involved. effective analysis and investigation of tax evasion groups is challenging for the following reasons. •the detection of tax evasion groups depends on the analysis of the topological interest relationship between different taxpayers and their various tax-related characteristics, which makes the investigation of tax evasion groups very complicated.•the ambiguity in the accounting principle has led to a time-consuming audit method for tax officers to manually check whether a suspected group is committing tax evasion or not. even recent advanced approaches to automatically detect group tax evasion can result in a high false positive rate and a large number of suspicious cases.•correlating and analyzing large volumes of financial data and transactions of related parties is time-consuming and complicates the exploration of suspicious patterns of tax evasion groups.methods and materialto address the above challenges, we try to use graph mining to help tax officers identify, extract, and explore suspected tax evasion groups. it should be noted that today, deep learning is considered one of the hot topics in the fields of machine learning, artificial intelligence, as well as data science and analytics. in addition, it has become a rapidly developing approach for research in the field of artificial intelligence. few methods in the literature can realize the advantages of machine learning-based audit methods and graph-based audit methods in detecting tax evasion based on related party transactions. in this study, we propose a new approach to effectively identify suspicious groups that exhibit both structural and commercial characteristics of tax evasion based on related party transactions through a uniform identification process. in this regard, information on net sales, cost of goods sold, operating profit (loss), special profit (loss), declared tax, and final tax from the performance tax returns of companies with related party transactions from 2016 to 2019 and the tax systems of the iranian tax affairs organization are used with confidentiality. in addition, data from 16,756 purchase and sale transactions have been extracted and used to monitor the pricing of related party transactions between parties. then, we identify behavioral patterns of tax evasion based on related party transactions by extracting structural features from the network of related party interest groups and theoretically deducing the business characteristics of tax evasion instruments based on evidence. results and discussionin this study, an innovative hybrid model based on deep learning networks and graph mining has been presented to help determine topological relationships. based on the results, both profit retention and profit transfer were observed in tax evasion of related party transactions. the results indicated that graph mining models are more accurate than the conventional logit and probit models. based on the results obtained, the intensity of the profit retention relationship in tax evasion of related party transactions is stronger than the profit transfer relationship. given that previous research in this area has not been conducted with this process, in the general trend, it can be stated that the present study is in line with the results of moqri gerdroudbari et al. (2023); ghanbarinejad et al. (2023); sedaghati et al. (2021); seyedhossein nasel mousavi et al. (2020); javadian kootnai et al. (2020); mohammad namazi et al. (2019); yating lin et al. (2023); leit et al. (2022); (2016); roan et al. (2019); tian et al. (2019); and gonzalez et al. (2013). conclusionbased on the results of the study, the following suggestions can be made. in order to increase the accuracy of the model, the iranian tax administration should use the information available in the databases of the companies registration office and the civil registration office to identify new interactive relationships and new behavioral topologies. the iranian tax administration should plan to identify complex and networked relationships between companies, which will play a key role in identifying suspicious groups and their tax evasion. the use of the aforementioned method by the iranian tax administration will update the monitoring systems and use advanced technologies to identify suspicious transactions. in addition, the aforementioned method can also be used to better monitor and supervise financial activities in free trade and special economic zones. given the phenomenon of tax evasion through transactions between related parties, related business units with specific tax purposes should be specifically defined. it is recommended that standard-setting bodies and regulators establish standards and rules to require companies to disclose more of the economic substance of related party transactions. investors are advised to use management contracts or audit mechanisms to reduce the detrimental effects of agency costs on related party transactions, by employing an audit committee in companies to control the price received or paid to related parties. it is recommended that the tax administration of the country, within the framework of the “risk-based tax audit” project and the comprehensive tax plan, and using the data available in tax systems and databases, take action to identify related party transactions and financial irregularities in them, and then take special measures to train the tax officers in charge to be more careful in the auditing process of related files. in this regard, transactions with related parties that are identified by the tax administration based on the information available in tax systems and databases should be considered as one of the criteria and indicators for determining the risk of economic operators in the process of selecting appropriate files for auditing taxpayers who are members of the “taxpayer system” or outside that system. penalties for failure of related business units to perform their duties, such as preparing and sending the related party transaction form to the relevant tax office, should be determined. an appropriate procedure should be developed to resolve disputes between taxpayers and tax authorities on issues related to related party transactions. to enact a clear and specific legal article on observing the arm’s length principle in financial transactions between related business units, and to develop precise tax guidelines and procedures, it seems necessary to conduct research related to transfer pricing before taking any action. it is suggested that the tax affairs organization actively cooperate with other countries that have experience in legislation, selecting appropriate cases for investigation, auditing, resolving disputes with taxpayers, and approving agreements on this matter. cooperation of the tax affairs organization with other countries to exchange data and information of taxpayers related to the issue of related party transactions in the form of agreements to avoid double taxation.