|Institution:||Delft University of Technology|
|Keywords:||payments; data mining; association rule mining; moral hazard; principal-agent problem|
|Full text PDF:||http://resolver.tudelft.nl/uuid:9768c196-e490-4ad7-b9e1-0912c5ebf6a4|
A significant amount of payments fail for seemingly ‘mysterious’ reasons when refused by an issuer (i.e. cardholder’s bank). The technical and social complexities, such as a principal-agent problem, hinder the development of policies or information architectures, which impose an issuer to share its ‘honest’ reason. In this thesis we propose a method to induce the authorisation decision rules which an issuer uses to refuse certain payments. We evaluate the method on a large database with payment transactions. Experts confirm the correctness and the completeness of the found rules. We experiment with strategies to act on the rules, which shows there is significant business value in the results. The results signal that there is a large potential which justifies further research to improve the method.