Usage of Artificial Neural Network and Support Vector Machine model for classification of Credit Scores

Abstract

In the emerging banking sector, credit is an important product. The decision to give or not to give credit to the customer is a decision that should be taken carefully from the point of view of the bank and as credit requests increase, the evaluation of applicants becomes even more complex. Decisions may be subjective because the evaluators may consider different criteria. In this case, various statistical and nonstatistical techniques are used to answer both the increasing number of applications and to make objective decisions without subjective criteria. In this study, we tried to distinguish between good and bad customers with twenty variables of the german loan data set, and the results of the applications are compared with one another. Some non-statistical techniques were used in the study: Artificial Neural Network and Support Vector Machine and the practice of these techniques are discussed. Practice presented as theoretical information without their practice are Logistic Regression, Random Forest, Decision Tree and K-Nearest Neighbor Approach. Practice related to these techniques will reprieve to work in the future. Since there are various advantages and disadvantages in the implementation of models, it can be said that the model with the highest prediction success, according to the data set used is Support Vector Machine -Vanilladot Kernel method.