Combination of CDM, ANFIS & MH Algorithms in a model to determine Fraudulent Financial-Tax Report

Abstract

Companies sometimes file fraudulent financial statements for tax fraud. The purpose of this study is to combine data mining tools and artificial intelligence with meta-heuristic algorithms to explain and optimize a model for detecting fraud and tax evasion by using the capacity of financial reporting. Qualitative and quantitative indicators of financial reports of 1056 year- companies in the Tehran Stock Exchange in the period of 2006 to 2019 were studied in the classical approach and used to expand the model in the Adaptive Neural-Fuzzy Inference System. Findings show that in optimization with genetic algorithm, particle swarm optimization algorithm and differential evolution algorithm, the most efficient model is obtained by particle swarm algorithm, which is the most efficient algorithm in the study with experimental and educational data. The results indicate that the application of different optimization algorithms in the data mining approach increases the predictive power of the fraudulent financial-tax reporting identification model

Keywords

• tax fraud
• financial reports
• adaptive neural-fuzzy inference system
• optimization algorithms