Are You Using Machine Learning to Improve Continuous Monitoring?
December 01, 2020
Are You Using Machine Learning to Improve Continuous Monitoring?Download Case Study
Find out how FTI Consulting’s Data & Analytics team used deep neural networks (artificial intelligence) to identify improper payments with 99% accuracy.
FTI Consulting was instructed to assist with an internal investigation which involved our Data & Analytics team building, training, and deploying an ensemble of deep learning neural networks (artificial intelligence) to identify potential improper payments and expenses.
We initially performed our investigations of potentially improper payments using a combination of manual processes and SQL scripts. A set of rules were defined to identify the improper payments/expenses based upon various data attributes and their respective correspondences or linkages across a procurement system (PS) and General Ledger (GL).
If a correspondence or linkage between PS and GL could be established based upon the rules, the relevant row(s) in each dataset were flagged by us as linked (i.e. potentially improper).
Once trained, our model consistently replicated the rules with over 99% accuracy. It also predicted potentially improper items in new datasets with high precision.