DSpace Arşivi :: by Yazar "Telimen, Mehmet" değerine göre listeleniyor

Listeleniyor 1 - 1 / 1

Predicting credit default risk using machine learning algorithm
(İstanbul Bilgi Üniversitesi, 2019) Telimen, Mehmet; Soybilgen, Barış
In this study, it was aimed to construct the analytical models that predict the probability of default of consumer credit by using machine learning algorithms. The data belonging to the customers of a bank has been used by making anonymity from the bank's test environment. This data set was composed of the lending status of the customers in the bank and the questioned credit bureau data at the credit application stage. Half of the samples in the data set were selected from those who had been in default and half were not. In the study, four of the widely used techniques of classifıcation based on machine learning have been discussed. Those are Logistic Regression, Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), and KNearest Neighbors (KNN). For each model, half of the data set was used for training and the other half was used for testing. Those models, which were trained with the same training set using the corresponding functions in R studio with R programming language, were tested with the same data set and the accuracy rates of them were compared. As a result of the comparison, with given this data, it is observed that the model of the Logistic Regression estimated the probability of default of the consumer loan with the highest accuracy rate which was 58.30%.