Internal Audit is responsible for providing the 3rd line of defense assurance over the effectiveness of controls in mitigating enterprise risks. We are primarily a judgment-based operation, relying on "humanness" to ascertain if risks are sufficiently being mitigated. This sort of environment makes it difficult to employ machine learning, given the ambiguity of decisions and the need for interpretability to back up decisions that were made. However, these limitations give us the ability to become more imaginative, finding unique ways to employ machine learning. In this talk, Andrew will provide two examples of prototypes being used in audit, an unsupervised machine learning exploratory "clustering" environment to provide insight into looking at data in new ways; and a supervised NLP model that classifies audit reports into different classes for use in reporting.
Andrew Clark is a Principal Machine Learning Auditor at Capital One where he is creating machine learning powered applications to reinvent the audit process. He is also establishing approaches for auditing and interpreting machine learning algorithms. His primary research focus is the application of advanced statistical and computational techniques to create value-added financial auditing solutions with the use of open source software, primarily in the Python ecosystem. Andrew received a B.S. in Business Administration with a concentration in Accounting, Summa Cum Laude, from the University of Tennessee at Chattanooga and an M.S. in Data Science from Southern Methodist University. He also holds the Certified Analytics Professional, American Statistical Association Graduate Statistician, and AWS Certified Solutions Architect - Associate certifications.