A Credit Scoring Model based on Strongly-Typed Genetic Programming

Rojin Aliehyaei and Shamim Khan


Credit scoring , genetic programming, modeling and evaluation


In this paper, we propose a credit scoring model for the efficient learning of rules to evaluate creditworthiness of credit applicants using a strongly-typed genetic programming approach. Our proposed model was implemented in Java under the framework of ECJ software and evaluated with the widely adopted German dataset. The experimental results demonstrated that the genetic programming approach is an efficient way for generating good solutions (or rules) to evaluate credit risk. The accuracy of rules discovered by GP ranged from 79% during training to 71% during testing.

Important Links:

Go Back