Predicting bonity of clients through two recursive partitioning methods
Language: en
Last modified: 2016-01-07
Abstract
The aim of the paper is to show and compare some classical classification approach with not so classical and typical classification method for identification of the factors influencing the credit scoring of some bank customers. For this purpose, we used statistical methodology like classification and regression trees (CART) and recursive partitioning method called PARTy. These classification methods are both not parametric methods and their application is not restricted by a strong normality assumption like LDA. Our approach is demonstrated on segment of the data coming from bank institution. Data pre-processing and the numerical computation were carried out in the programming language R.
Keywords
classification; credit scoring; CART; PARTY
Full Text:
PDF
