Statističke metode kreditnog skoringa

Kovačević, Marija (2014) Statističke metode kreditnog skoringa. Diploma thesis, Faculty of Science > Department of Mathematics.

[img] PDF
Restricted to Repository staff only
Language: Croatian

Download (465kB) | Request a copy

Abstract

Credit scoring is a statistical method that is used to assessing the risk that a new client who applies for a loan, or an existing client delay in repayment of the loan or it will not even be able to pay off the loan. In other words, credit scoring is a method for determining the credit risk of carrying a client. Using historical data and statistical credit scoring techniques trying to isolate the effects of particular features that a customer lead to a situation where they can not repay the loan. The result of this method is the score that bank used to rate its customers based on the risk that they carry. Based on that, how much risk the bank wants to accept it, is determined by marginal score, so that customers who have a higher score than the border will be granted credit and those who have less imminent be rejected. To create the model, analysts analyze historical data previously approved loans, ie. analyze how the previous customers behave when repayment of the loan, in order to determine the characteristics that are useful in assessing whether the client is capable of repaying the loan regularly. Therefore customers are divided into good and bad. The objective is to accurately classify customers into good and bad, because misclassification brings higher costs that arise. If a customer who has a good classified as poor, thus the client will be denied and are lost profits that could be realized that he approved credit. However, a much larger error to the client, which is classified as bad is good and grant him credit for time incurred certain losses when the client is no longer able to repay the loan. So we have a very essential that these errors to be minimized. However, it is not so simple. Reducing errors resulting misclassification of bad customer leads to an increase in errors of misclassification good customers and vice versa. For this reason, you should consider what is the relationship of losses resulting from the incorrect classification on the basis of the criterion to decide which errors should be minimized.

Item Type: Thesis (Diploma thesis)
Supervisor: Slijepčević, Siniša
Date: 2014
Number of Pages: 35
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 03 Jun 2015 11:33
Last Modified: 23 Jan 2017 06:39
URI: http://digre.pmf.unizg.hr/id/eprint/3985

Actions (login required)

View Item View Item