An Expected Value Model for Maximizing Revenues from Sales Tax Audits

Building on our previous project, my colleagues and I revisited the problem of maximizing the revenue of a tax auditing operation. We conducted further research, restructured the problem and the evaluation criteria to better resemble real-world auditing situations, and revisited our model.

Our new model, created by combining a more rigorous regression with a bagged decision tree, gave an even better result. Our research showed our solution was unpublished, outperformed other methods, and was simpler to execute than published methods.

Our research was accepted by the University of Texas' Workshop on Social & Business Analytics, and we made a poster intentionally targeted towards the business-focused audience:

We are now considering some further refinements, with the goal of eventually submitting the work for publication.