A Naive Bayes approach to fraud prediction in loan default
Ibukun Eweoya, Ayodele Ariyo Adebiyi, Ambrose Agbon Azeta, Felix Chidozie, Frank Agono, Blessing Guembe · Journal of Physics Conference Series · 2019
Abstract The essence of granting loans to individuals and corporate beneficiaries is to boost the economy while the lenders make profit from the interest that accrues to the lending. However, due to non-compliance to basic rules, fraud is prevalent in credit administration and traditional methods of detecting fraud have failed. Furthermore, they are time-consuming and less accurate. This work uses a supervised machine learning approach, specifically the Naïve Bayes to predict fraudulent practices in loan administration based on training and testing of labeled dataset. Previous works either predict credit worthiness or detect loan fraud but not predicting fraud in credit default. The approach employed in this work yielded 78 % accuracy.