Qualification of the bank costumers with SAS on the base of bank products

OData support
Supervisor:
Györfi László Dr.
Department of Computer Science and Information Theory

ABSTRACT

The financial institutions offer a wide range of credits for the members of the private and

economic sectors. In order to monitor a client and to form a valid position about solvency and

reliability, there is need for a mature, accurate client-qualifying system. The retail credit section has

been developing since 2003 in Hungary, similarly to the western countries, since then we are using

modern client-qualifying systems, but these applications don’t always give satisfactory results.

During a work at a medium-sized bank in Hungary, I have met an applied classification system,

which did not rank the clients well in every case. In order to solve this problem, I chose the

qualification of the bank clients as the topic of my thesis. I tried to work out a more accurate method

in order to make a better classification about the bank clients. I used the SAS softwares to solve the

computer science tasks.

In my thesis, I continued to improve a bank client-qualifying system in order to be able to predict

the clients more efficiently in terms of their payment willingness, and to filter out the potential bad

clients before the credit disturbement. In this methodology, I assign scores to the clients on the basis

of their activities from the viewpoint of the bank.

As a first step, I’m presenting the currently applied system, then I’m revealing my suggested

methodology. Through an actual example I will apply the adapted and improved algorithm that

qualifies the clients more efficiently. I generated sample data, since I might not have used real bank

data for the comparison of the algorithms.

The main aspect of realizing the suggested system was not to score the events themselves

separately, but to find the most suitable score point by looking into the context of these events. After

the forming of the algorithm, I obtained the scores to both systems. Then I made the assessment of

the results, first with the help of elemental statistics, then with the Credit Scoring method. The final

results of the analysis indicated that the new methodology is more efficient than the actual method

applied by the bank.

I think I managed to prepare an algorithm which can be applied better, one that was tested with

scoring analysis. I would suggest my methodology for other banks, because it demonstrates that I

worked out a better client-qualifying system that can filter out the insolvent clients more efficiently.

Downloads

Please sign in to download the files of this thesis.