Invoicing is a useful tool that facilitates large parts of our economic system, but can also be of great risk to corporations when debts are left unpaid. Summoning someone to court is a costly procedure to force them to pay after all. However, when this person is simply unable to pay, this forms a significant loss for business and society. This thesis explores the decision making process behind summoning someone, looking to predict whether it is favourable to start these proceedings. We train a wide range of models from a statistics and machine learning background to perform this analysis. Various model evaluation measures show that a Random Forest is of best use, providing a higher increase in performance than is typically seen in closely related fields like credit scoring. The results obtained from the model, albeit low according to conventional standards, allow us to identify a sizeable share of potential cases that are better pursued outside of a courtroom. Implementation of the model could reduce investments into the summons process by 10 to 20 percent, proving to be of great business value.

Additional Metadata
Keywords Keywords: Credit Management, Summons Process, Machine Learning, Predictive Modelling, Random Forest.
Persistent URL hdl.handle.net/2105/47331
Series Econometrie
Citation
Koomen, Y. de, & Birbil, S.I. (2019, April 30). Increasing Summons Efficiency with Predictive Modelling. Econometrie. Retrieved from http://hdl.handle.net/2105/47331