The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9687

Title:Credit Scoring Using Neural and Evolutionary Techiques
Date: 1996
Abstract:In today's world credit is seen as an essential component of daily life which explains the enormous growth in credit card ownership over the past twenty years. As a consequence of this growing demand, creditors want to increase their revenues by granting credit to more creditworthy applicants and at the same time they want to reduce their loses by refusing credit to noncreditworthy applicants. Credit Scoring is a technique used by credit grantors for assessing the creditworthiness of an applicant in order to aid the credit granting decision for a loan or a credit card. In this project the predictive power of neural networks, decision trees and genetic algorithms in credit scoring for credit cards is investigated. All these techniques learn from examples provided to them and they can capture non-linear relationships which make them alternative to the conventional techniques such as discriminant analysis and regression. The results show that traditional techniques compare very well with the techniques studied in this project. Neural Networks performed slightly better than Genetic Algorithms and Decision Trees.

[Search These Pages] [DAI Home Page] [Comment]