Machine Learning Algorithms for Detecting Tax Fraud: Application and Challenges in Nigeria

Abstract

The study focused on machine learning algorithms for detecting tax fraud especially its application and challenges in Nigeria. To achieve this, survey research design was adopted. The population of the study is made up of 150 participants, including Data Analysts, Financial Accountants, Tax Consultants and Tax Auditors, who were chosen to offer a variety of perspectives on the practices of detecting tax fraud. The study adopted convenience sampling techniques. 150 questionnaire instruments were distributed evenly across the participants but only 115 were correctly completed and returned. The study used a descriptive field survey research design and adapted questionnaire as a research instrument. The data collected were analyzed using both descriptive statistics and simple regression analysis. The findings of objective one indicates that tax fraud is a significant issue in the Nigerian tax system, and that weak enforcement of tax laws and corruption among tax officials are the major causes of tax fraud in Nigeria. The findings from the analysis of objective two revealed that although machine learning techniques are seen to be effective tools for improving fraud detection, structural barriers hinder their broad implementation. These include limited access to quality data, ethical concerns, institutional resistance, insufficient government funding, and lack of regulatory clarity. Finally, the findings from the analysis of objective three demonstrate a strong and statistically significant effect of ML algorithms on improving tax fraud detection (R² = 0.899, p < 0.001), indicating that, when applied properly, technology adoption may significantly reduce fraud. Based on the findings, the study recommends that to guarantee that ML algorithms can operate with dependable input, the Nigerian government should make investments in digitizing tax records and enhancing data quality. The study also recommends that clear guidelines should be established to control the ethical use of AI in tax systems, with special emphasis to transparency, accountability, and privacy.