Management Science Lunchtime Seminar: 'Genetic algorithm derived models for optimisation of alternative measures of perf
Tuesday 17 January 2006, 13:00
LT9 (B98), Management School
Using Genetic Algorithms to Develop Scoring Models for Consumer Credit Assessment
Steve Finlay
Tuesday 17th January, 1300-1400
LT9 (B98), Management School
Note that a buffet lunch with foods and drinks (orange juice, coffee, tea) will be provided. To be able to order the right amounts of food, please let Gay know if you are planning to attend by sending an email reply by the Friday preceding the seminar. Of course, feel free to bring your own lunch if you so prefer.
Abstract: Most approaches to credit scoring generate the parameters of a model by minimising some function of individual error, or by maximising likelihood. In practice, the criteria by which the parameters of a model are determined and the criteria by which models are assessed may differ. Practioners tend not to be interested in standard statistical measures of model fit such as the R2 coefficient for linear regression or the likelihood ratio for logistic regression. Instead, performance will often be assessed using global measures such as the GINI coefficient, or by considering the misclassification rate at different points in the distribution of model scores. In this paper an approach using genetic algorithms is described, where the training algorithm is used to directly maximise/minimise the performance measure of interest. Empirical results are presented that show that the GA approach has the potential to generate scoring models that are competitive compared to a range of models constructed using more traditional approaches.
