Modeling mortgage assessment with computational argumentation theory and defeasible reasoning

Document Type



This item is available under a Creative Commons License for non-commercial use only


Computer Sciences

Publication Details

A dissertation submitted in partial fulfillment of the requirements to the Technological University Dublin for the degree of M.Sc. in Computing (Stream: Advanced Software Development)


In the mortgage lending business of a bank, a key focus area is risk analysis, which supports the mortgage awarding process and the prediction of the risk of defaulting (repayment issues). The standard risk assessment method at most banks is a scorecard calculation. A new way of predicting the defaulting is proposed, which has not been done before, using Defeasible Reasoning (DR) and computational Argumentation Theory (AT), which are areas of interdisciplinary research, in the discipline of Articial Intelligence (AI). Argumentation is formalised by reasoning models which are inspired by human reasoning. For a more realistic representation AT employs DR which is a non-monotonic reasoning process, meaning that in case of new evidence a previous conclusion might change. The computational AT approach is predominantly knowledge driven and it includes building and evaluating arguments, their relationships, the resolution of their inconsistencies and the generation of defeasible conclusions, on which the experiment conducted in this thesis is based upon. It is demonstrated how is it possible to reasoning on a defeasible way to predict the risk of defaulting. Results demonstrated that in 75% of the test cases AT predicted better. The prediction accuracy was compared using a so-called confusion-matrix (2 by 2 squares) where the two axes being Predicted (Yes; No) and Actual (Yes;No) and a limitation applied as only half of the confusion-matrix could be used, due to unavailable and unverifiable data.

This document is currently not available here.