MSU paper describing a real-life use of an expert system paired with an evolutionary training algorithm in the realm of US credit underwriting.

The original paper is linked below. Some sections were co-authored with Rym Oulad Ali (ENSEA/MSU).

In this experiment, we implement a CLIPS fuzzy expert system to underwrite consumer loans. It is then tested against a real-life 2023 portfolio from Texas. Afterwards, I propose and implement a genetic algorithm to tune the parameters of our CLIPS code. Combining these two technologies results in a very performant, fully automated underwriting algorithm.