Underpricing Fraud Detection using Artificial Intelligence Technology to reduce Tax evasion in Tanzaniaa

The dataset includes both qualitative and quantitative data collected during research on the prevalence of tax evasion through product underpricing. This dataset is designed to be suitable for machine learning applications to detect underpricing fraud in real time. The qualitative data can be analyzed using Atlas.ti software, while the quantitative data is compatible with SPSS. The qualitative data was gathered from traders in the Arusha region, with ethical clearance from the Nelson Mandela African Institute of Science and Technology and informed consent from participants. Google Forms was used to distribute questionnaires to those with internet access, while in-person visits were made to businesses for participants without smartphones or computers. No data was collected from individuals who declined to participate, adhering strictly to ethical guidelines. All collected data is anonymous, except for publicly available information. Interviews were conducted with tax authority officials based at the headquarters in Dar es Salaam, including officers from the EFD department, ICT officers, and economists. These interviews provided insights into the extent of tax evasion in Tanzania, existing mechanisms to combat it, persistent loopholes, and opinions on proposed measures to address underpricing fraud. Additionally, document analysis was performed on publicly available government reports to assess trends in tax collection. Media sources, including images, videos, and magazines, were also analyzed qualitatively to better understand the intensity of tax evasion, the challenges the government faces, and the efforts made to boost revenue.