Journal of Tax Research

other فرار مالیاتی اشخاص حقوقی با استفاده از هوش مصنوعی Detecting Tax Evasion of Legal Entities Using Artificial Inteligence مدیریتی Management پژوهشي Research فرار مالیاتی یکی از دغدغه‌های مستمر نظام‌های مالیاتی بخصوص در کشورهای در حال توسعه می‌باشد. پدیده فرار مالیاتی از این جهت قابل تأمل است که ضمن کاهش درآمدهای دولت و افزایش سطح شکاف مالیاتی نسبت درآمدهای مالیاتی به تولید ناخالص داخلی را تنزل می‌دهد. از آنجا که فناوری داده‌کاوی از قابلیت‌های پیش‌بینی و طبقه‌بندی فراوانی برخوردار است می‌تواند فرآیند تصمیم‌گیری در مسائل مالی را تسهیل نماید. لذا در این تحقیق سعی کردیم از الگوریتم‌های هوش مصنوعی همچون الگوریتم‌های شبکه عصبی مصنوعی، ماشین بردار پشتیبان و الگوریتم کلونی زنبور مصنوعی جهت تشخیص فرار مالیاتی استفاده کنیم. جامعه آماری پژوهش حاضر شامل 3600 مؤدی مالیاتی در سال‌های مورد مطالعه (1381-1391) استان تهران می‌باشند. بنابراین با استفاده از الگوریتم‌های ترکیبی در این تحقیق توانستیم با دقت 85.59 درصد برای داده‌های آموزشی و دقت 83.79 درصد برای داده‌های آزمایشی به تشخیص و طبقه‌بندی مودیان متلقب در پرداخت مالیات و مودیان با پرداخت به&lrm;موقع مالیات بپردازیم. بنابراین با توجه به نتایج بدست آمده در می‌یابیم که روش پیشنهادی در این تحقیق از توان بالایی در شناسایی مودیان متقلب از پرداخت مالیات برخوردار است.   By using artificial intelligence methods, tax evasion of legal entities which is a continuous concern of tax system, has been detected based on financial data recorded in financial transactions of collections. The phenomenon of tax evasion can be considered in the sense that it reduces the ratio of tax revenues to GDP while reducing government revenues and increasing the level of the tax gap. Since data mining technology has many predictive and classification capabilities, it can facilitate the decision-making process in financial matters. The different methods for transaction data modeling are based on matching labeled data and the reviews of legal taxpayers' files by tax experts, which are marked as tax/non-tax evasion cases. There are various methods for modeling on this method. With the increasing use of artificial intelligence modeling, this article, by using three modern methods of artificial intelligence, on the aforementioned data set, validates the use of these methods by using the parameters of accuracy, sensitivity, and precision, which determine the quality of the produced model from various aspects. It validates, the feasibility of using these models has also been done on local financial transaction data. Therefore, in this research, it has been tried to use artificial intelligence algorithms such as artificial neural network algorithms, support vector machine and artificial bee colony algorithm to detect tax evasion. The statistical population of the current research includes 3600 taxpayers in the studied years (2002-2012) of Tehran province. One of the new methods of financial data modeling is the use of hybrid methods, in which the combination of several artificial intelligence methods is used to increase the quality of the model. It is obvious that the combination of the mentioned methods has favorable results in some situations and may not be so in other situations. Therefore, in this article, this issue has been investigated in detail and the use of different methods and a combination of artificial intelligence methods has been discussed. Therefore, by using the combined algorithms in this research, we were able to with an accuracy of 85.59%; For training data and accuracy of 83.79% for test data, let's identify and classify taxpayers who are late in paying taxes and taxpayers who pay taxes on time. Therefore, according to the obtained results, we find that the proposed method in this research has a high ability to identify taxpayers who cheat on paying taxes. The use of artificial intelligence systems can be used as an economic stimulus in order to reduce tax evasion. In this regard, it is suggested to first form an integrated system as an economic data center, and this center will take on the responsibility of keeping the country's financial data safe. Then, a standard for the use of artificial intelligence and data mining of financial data should be developed. With this method of scientific and academic research, without providing sensitive data to outside the required organizations, it is possible to take a step forward from the scientific ability of experts regarding factors such as tax evasion detection, terrorism financing detection and money laundering in order to improve the country's economic system.   فرار مالیاتی, هوش مصنوعی, اشخاص حقوقی, داده کاوی Tax Evasion, Artificial Inteligence, Legal Entities, Data Mining 200 217 http://taxjournal.ir/browse.php?a_code=A-10-2433-2&slc_lang=other&sid=1 Hamidreza Rostam Beiggi حمیدرضا رستم بیگی krayeneband@ihu.ac.ir 100319475328460010423 100319475328460010423 No Meghdad Ayneband مقداد آینه‌بند Me.Aynehband@iau.ac.ir 100319475328460010424 100319475328460010424 Yes