Author/Authors :
Sadatrasoul، Seyed Mahdi نويسنده , , Gholamian، Mohammad Reza نويسنده , , Hajimohammadi، Zeynab نويسنده Amirkabir University of Technology, Department of computer science, Tehran, Iran , , Hosseini، Mahdi نويسنده Advanced Information Systems Research Group, ICT Research Institute of ACECR, Tehran, Iran ,
Abstract :
Credit scoring mainly distinguishes good customers from the bad ones; therefore it is a classification problem. There are many techniques introduced to solve the problem such as support vector machines, neural networks and rule based classifiers. The main objective of this process is to maximize the profit of bank or financial institute. However these traditional methods of classification seem not to support this objective well. This paper investigates this issue and shows that the best classification model is not necessarily the most profitable model. The applications of the models are shown on an ironing real credit dataset since 2007 to 2012.
