Volume 1395, Issue 77 (2016)                   J Tax Res 2016, 1395(77): 0-0 | Back to browse issues page

XML Persian Abstract Print


Abstract:   (4649 Views)

Abstract

Tax fraud includes a large spectrum of methods including denying the facts and realities, claiming wrong information and performing financial businesses without considering legal frameworks. Nowadays, with the development of tax systems and the large volume of tax data, it is necessary to have tools to process this large data and to exploit information and knowledge. According to tax policies, especially in value-added tax resource, the rate of tax fraud is increasing. Based on the investigations, researchers use standard methods such as association rules, clustering, neural networks, decision trees, Bayesian networks, regression and genetics to detect tax fraud. Because of large volume of tax database, most of the studied algorithms are time consuming. At first, Apriori Algorithm was used. This algorithm was one of the unsupervised learning models and association rules. It is used to detect suspicious behavior of tax fraudsters. Secondly, a system for tax fraud detection based on Bayesian networks is presented and its performance is improved using parallel processing techniques. Results of the study show that using available parallel processing patterns improve the execution time of tax fraud detection algorithm considerably.

Full-Text [PDF 757 kb]   (2724 Downloads)    
Type of Study: Research | Subject: Economic
Received: 2016/08/15 | Accepted: 2016/08/15 | Published: 2016/08/15

Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.