ID: 18126
Authors:
Maria Teresinha Arns Steiner, Celso Carnieri, Bruno H. Kopittke, Pedro J. Steiner Neto.
Source:
RAUSP Management Journal, v. 34, n. 3, p. 56-67, July-September, 1999. 12 page(s).
Keyword:
neural networks , pattern recognition , probabilistic expert systems
Document type: Article (Portuguese)
Show Abstract
Recognising and foreseeing which credit customers will be "good or bad payers" is an important and difficult task to banking institutions and to credit protection services. With 2.855 historical registers of a German bank, we studied, in this work, the Probabilistic Expert Systems and Neural Networks techniques, both in the Artificial Intelligence area, comparatively, using the shell SPIRIT and MatLab-Neural Networks Toolbox. These techniques allow us to recognise patterns as well as their further use in later evaluations. The evaluation of a given customer and the probability of the payment of a loan leads to a risk/payment rate for the bank, based upon interest rates, loan amount, payment period and other fees, establishing a safe criteria to determine when to concede credit.
