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 fulfilment of the requirements of Technological University Dublin for the degree of M. Sc. in Computing (Data Analytics) May 2014.


The management of credit risk and mortgage arrears has become a very important area in financial services and banking. This dissertation set out to build a statistical model, which incorporates customer spending habits and the current equity value of a property, capable of predicting arrears. Current literature identifies many themes such as negative equity and unemployment that are common occurring factors in mortgage arrears but a multi-faceted approach was required to build a model capable of accurately predicting arrears. Property equity values were included in the model by taking the current outstanding value of the loans and using a property price index to work out the current market value of the property. Transactional data was included in the model as an indicator of the spending habits and trends of the borrowers by deriving monthly values for savings, discretionary spend, necessary living expenses and mortgage payments to give an indication of their overall financial health. Different modelling techniques were applied to the data along with numerous sampling methods in an endeavour to achieve the best results. The models were evaluated using misclassification costs as well as the more traditional measures such as recall, specificity, precision and overall accuracy. The created models achieved a high level of accuracy in predicting arrears a number of months in advance. Even though much of the existing literature on predictive models for mortgage arrears and default favours the use of Neural Networks for this type of classification, it has been shown here that Decision Trees achieved both the better and most consistent scores. The resulting models created achieved a high level of accuracy and were capable of predicting arrears a number of months in advance of the arrears actually occurring. The overall experiment was a success, and it proved that transactional data and equity values can help to improve the accuracy and predictability of an arrears model.