Corporate Tax Compliance and Fraud Prevention System by Principal Component Analysis and Auto-Encoder

Almost all tax authorities prioritize the detection of tax fraud as a means of optimizing their earnings and upholding strict compliance standards. When devising standard tax models and their outcomes, it is imperative for every nation to take into account the prevalence of tax evasion and avoidance, as they can significantly skew tax arrangements. Early models in this field focused on tax evasion, as it was perceived as a challenge to the state’s ability to enforce the law. This study expands on earlier research on tax fraud detection, which used a variety of machine learning models to look at all tax kinds. This model finds a number of variables that are important for identifying income tax fraud in both small and corporate enterprises, such as the length of the firm, its size, and whether it works locally or internationally. Notably, this study compares several parameters including taxable income, tax rate, deductions to discover the best settings for obtaining improved accuracy. It is the first to use Principal Component Analysis and Auto-Encoder to detect corporate tax fraud. The findings of this work will assist scrutinizers in comprehending the elements that lead to corporate tax fraud, hence decreasing the duration and expenses of audits and facilitating the recovery of lost revenue from corporate tax fraud.