Qualitative Modelling of Credit Scoring: A Case Study in Banking
Several modelling procedures have been suggested in the literature that aim to help credit granting decisions. Most of these utilize statistical, operational research and artificial intelligence techniques to identify patterns among past applications, in order to enable a more well-informed assessment of risk as well as the automation of credit scoring. For some types of loans, we find that the modelling procedure must permit the consideration of qualitative expert judgements concerning the performance attractiveness of the applications. In this paper, we describe in detail the various steps taken to build such a model in the context of the banking sector, using the macbeth interactive approach. The model addresses the scoring of medium and long term loans to firms, to enable the multicriteria assignment of each application to a category which may range from rejection to acceptance with different spreads.