نوع مقاله : مقاله پژوهشی
نویسندگان
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Empowering needy people through the payment of employment facilities is one of the challenges and strategies of developed countries. Paying facilities to needy people due to their special characteristics with common customers of financial institutions has a high credit risk. Common models of credit risk analysis such as credit score cards, decision trees, and gradient boosting have generally been developed for small samples and little data. They do not provide volume.On the other hand, collecting customer data is a difficult task due to the laws of maintaining the security of information. According to the above problem, in this research, using adversarial generative neural networks, a model has been presented that can perform credit risk analysis with high accuracy even for small samples. In this model, which is based on The theory of games is that new data are generated by the neural network until the Nash equilibrium point is reached, and then the data generated with the same distribution is combined with the real data. The results of the proposed model compared to the developed gradient boosting method show It is that while overcoming the problem of lack of successful samples, the accuracy of customer credit risk analysis has increased by 32.7%.
کلیدواژهها [English]