Finding an Optimal Approval Strategy in Credit Scoring System Using Genetic Algorithm

The purpose of this study is to build a system that automatically generates an optimal strategy (AGOS) which can solve the business problem of maximizing the acceptance rate without taking more risks in loan approval process. This paper formulates the finding an optimal approval strategy of a credit decision as a constrained optimization problem. Genetic algorithm is used to solve it and tested on a real data from a South Korean online lending company. In our problem setting, the optimal solution consists of a set of cut-off thresholds and selected features, which later used to decide whether to accept the loan application or not. Adaptive ability of our AGOS system is investigated using simulated data that reflects different scenarios under dataset shift. The result shows that the developed system outperforms the current existing methods in terms of acceptance rate at the same risk level