This project was done for a huge credit bureau in the U.S
Traditional neural networks are typically “black boxes”, with the inability to explain the modeling outcome. NeuroDecision uses neural networks to evaluate risk predictions, while providing reason codes that allow businesses to meet regulatory requirements.This is how Equifax is able to deliver explainable neural network technology.
I worked on the Generation 3 of this algorithm for which a patent was issued on Feb 11 2020
I worked on modifying the objective function to implement the constrained solution in Apache Spark by extending the MLLib library