engleski

Credit scoring using robust logistic regression

The objective of this paper is to investigate the possibility of improving a predictive power of the credit scoring model using robust logistic regression instead of the classical one. The robust procedures are usually used for data sets containing outliers. As we deal with small business credits and a lot of non-financial variables, we cannot expect a data set without outliers. Some of the proposed robust procedures can be computationally very expensive. Our aim is to answer the question which robust procedure is to be used in order to get satisfactory predicting power and computational time.

credit scoring; robust method; logistic regression

nije evidentirano