Problems of building Bayesian networks for credit rating?

Question

fuzzy logic

Hey!

I need to justify the use of fuzzy logic to build an expert system for assessing loans.
I have been trying to understand Bayesian networks for a long time, and most importantly, to find a justification that they are not very suitable for assessing creditworthiness.

so far I have found such problems:
- Bayesian networks model dependencies in the process under study
- each BS variable is assigned a probability distribution. this is an Np-hard problem. Leading to problems: labor-intensive introduction of probabilities, large amount of storage

In 1994, Kosko proved the fuzzy approximation theorem: Any mathematical model can be approximated (with varying degrees of accuracy) using a fuzzy logic system
how to tie it to the rationale for the possibility of using fuzzy logic instead of BS,

thanks for the comments and help

Answer 1

- Bayesian networks model dependencies in the process under study

and scoring is not the summation of dependencies? ))

how to tie it to the rationale for the possibility of using fuzzy logic instead of BS

just build a model and show that it is equivalent to Bayesian,
you can for the main parameters = there are a dozen of them: gender, age, place, income, employment, spouse, property