Optimizing debt collections using constrained reinforcement learning

In this paper, we propose and develop a novel approach to the problem of optimally managing the tax, and more generally debt, collections processes at financial institutions. Our approach is based on the framework of constrained Markov Decision Process (MDP), and is unique in the way it tightly couples data modeling and optimization techniques. We report on our experience in an actual deployment of a tax collections optimization system based on the proposed approach, at New York State Department of Taxation and Finance. We also validate the competitive advantage of the proposed methodology using other data sets in a related application domain. Categories and Subject Descriptors I.2.6 [Artificial Intelligence]: Learning General Terms Algorithms Keywords Constrained Markov Decision Process, Reinforcement Learning, Debt collection, Business analytics and optimization

• Type: Conference
• Year: 2010
• Where: KDD
• Authors: Naoki Abe, Prem Melville, Cezar Pendus, Chandan K. Reddy, David L. Jensen, Vince P. Thomas, James J. Bennett, Gary F. Anderson, Brent R. Cooley, Melissa Kowalczyk, Mark Domick, Timothy Gardinier
• Keywords: Constrained Markov Decision, Data Mining, KDD 2010, Keywords Constrained Markov, Markov Decision Process