Scoringa modeļi fizisku personu kredītriska novērtēšanai
Loading...
Date
Authors
Advisor
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Language
N/A
Abstract
Maģistra darbs tiek veltīts kredītu pieteikumu Scoringa modeļa veidošanai. Darbā apskatīti visi Scoringa modeļa izveides posmi - sākotnējo datu sagatavošana un pārveidojumi, noraidīto pieteikumu slēdziens, mainīgo atlase modelī, parametru novērtēšana un iegūtā modeļa kvalitātes novērtējums.
Izmantojot loģistisko regresiju un neironu tīklus, tika izveidoti divi dažādi Scoringa modeļi un salīdzināta to kvalitāte.
Veidojot modeļus tika salīdzinātas divas pieejas datu sagatavošanā - fiktīvo mainīgo pieeja un liecības svaru pieeja.
This paper is devoted to the credit application Scoring model development. Paper describes all stages of Scoring development process - data preparation and transformation, rejected inference, variable selection, parameter estimation and model validation. Using logistic regression and neural networks two different Scoring models were developed and their classification efficiency have been compared. In development process two data transformation approaches were compared - dummy variable and weight of evidence.
This paper is devoted to the credit application Scoring model development. Paper describes all stages of Scoring development process - data preparation and transformation, rejected inference, variable selection, parameter estimation and model validation. Using logistic regression and neural networks two different Scoring models were developed and their classification efficiency have been compared. In development process two data transformation approaches were compared - dummy variable and weight of evidence.